Merge branch 'hotfix/dirac-ITT-fix'

Code stalls and resumes on KNL at cambridge. Curious.

CG iterations 23ms each, then 3200 ms pauses. Mean bandwidth reports
as 200MB/s. Comms dominant in the report. However, the time behaviour suggests it
is *bursty*.... Could be swap to disk?

.gitignore

@@ -9,6 +9,7 @@
 ################
 *~
 *#
+*.sublime-*
 
 # Precompiled Headers #
 #######################
@@ -47,7 +48,9 @@ Config.h.in
 config.log
 config.status
 .deps
-*.inc
+Make.inc
+eigen.inc
+Eigen.inc
 
 # http://www.gnu.org/software/autoconf #
 ########################################
@@ -89,6 +92,7 @@ build*/*
 #####################
 *.xcodeproj/*
 build.sh
+.vscode
 
 # Eigen source #
 ################
@@ -101,4 +105,21 @@ lib/fftw/*
 # libtool macros #
 ##################
 m4/lt*
-m4/libtool.m4
+m4/libtool.m4
+
+# github pages #
+################
+gh-pages/
+
+# Buck files #
+##############
+.buck*
+buck-out
+BUCK
+make-bin-BUCK.sh
+
+# generated sources #
+#####################
+lib/qcd/spin/gamma-gen/*.h
+lib/qcd/spin/gamma-gen/*.cc
+

.travis.yml

@@ -7,64 +7,8 @@ cache:
 matrix:
   include:
     - os:        osx
-      osx_image: xcode7.2
+      osx_image: xcode8.3
       compiler: clang
-    - compiler: gcc
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-4.9
-            - libmpfr-dev
-            - libgmp-dev
-            - libmpc-dev
-            - libopenmpi-dev
-            - openmpi-bin
-            - binutils-dev
-      env: VERSION=-4.9
-    - compiler: gcc
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-5
-            - libmpfr-dev
-            - libgmp-dev
-            - libmpc-dev
-            - libopenmpi-dev
-            - openmpi-bin
-            - binutils-dev
-      env: VERSION=-5
-    - compiler: clang
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-4.8
-            - libmpfr-dev
-            - libgmp-dev
-            - libmpc-dev
-            - libopenmpi-dev
-            - openmpi-bin
-            - binutils-dev
-      env: CLANG_LINK=http://llvm.org/releases/3.8.0/clang+llvm-3.8.0-x86_64-linux-gnu-ubuntu-14.04.tar.xz
-    - compiler: clang
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-4.8
-            - libmpfr-dev
-            - libgmp-dev
-            - libmpc-dev
-            - libopenmpi-dev
-            - openmpi-bin
-            - binutils-dev
-      env: CLANG_LINK=http://llvm.org/releases/3.7.0/clang+llvm-3.7.0-x86_64-linux-gnu-ubuntu-14.04.tar.xz
       
 before_install:
     - export GRIDDIR=`pwd`
@@ -73,13 +17,15 @@ before_install:
     - if [[ "$TRAVIS_OS_NAME" == "linux" ]] && [[ "$CC" == "clang" ]]; then export LD_LIBRARY_PATH="${GRIDDIR}/clang/lib:${LD_LIBRARY_PATH}"; fi
     - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then brew update; fi
     - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then brew install libmpc; fi
-    - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then brew install openmpi; fi
-    - if [[ "$TRAVIS_OS_NAME" == "osx" ]] && [[ "$CC" == "gcc" ]]; then brew install gcc5; fi
     
 install:
     - export CC=$CC$VERSION
     - export CXX=$CXX$VERSION
     - echo $PATH
+    - which autoconf
+    - autoconf  --version
+    - which automake
+    - automake  --version
     - which $CC
     - $CC  --version
     - which $CXX
@@ -92,15 +38,9 @@ script:
     - cd build
     - ../configure --enable-precision=single --enable-simd=SSE4 --enable-comms=none
     - make -j4 
-    - ./benchmarks/Benchmark_dwf --threads 1
+    - ./benchmarks/Benchmark_dwf --threads 1 --debug-signals
     - echo make clean
     - ../configure --enable-precision=double --enable-simd=SSE4 --enable-comms=none
     - make -j4
-    - ./benchmarks/Benchmark_dwf --threads 1
-    - echo make clean
-    - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then export CXXFLAGS='-DMPI_UINT32_T=MPI_UNSIGNED -DMPI_UINT64_T=MPI_UNSIGNED_LONG'; fi
-    - ../configure --enable-precision=single --enable-simd=SSE4 --enable-comms=mpi-auto
-    - make -j4
-    - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then mpirun.openmpi -n 2 ./benchmarks/Benchmark_dwf --threads 1 --mpi 2.1.1.1; fi
-    - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then mpirun -n 2 ./benchmarks/Benchmark_dwf --threads 1 --mpi 2.1.1.1; fi
-
+    - ./benchmarks/Benchmark_dwf --threads 1 --debug-signals
+    - make check

Makefile.am

@@ -1,5 +1,17 @@
 # additional include paths necessary to compile the C++ library
-SUBDIRS = lib benchmarks tests
+SUBDIRS = lib benchmarks tests extras
+
+include $(top_srcdir)/doxygen.inc
+
+bin_SCRIPTS=grid-config
+
+
+.PHONY: bench check tests doxygen-run doxygen-doc $(DX_PS_GOAL) $(DX_PDF_GOAL)
+
+tests-local: all
+bench-local: all
+check-local: all
 
 AM_CXXFLAGS += -I$(top_builddir)/include
+
 ACLOCAL_AMFLAGS = -I m4

README

@@ -1,44 +0,0 @@
-This library provides data parallel C++ container classes with internal memory layout
-that is transformed to map efficiently to SIMD architectures. CSHIFT facilities
-are provided, similar to HPF and cmfortran, and user control is given over the mapping of
-array indices to both MPI tasks and SIMD processing elements.
-
-* Identically shaped arrays then be processed with perfect data parallelisation.
-* Such identically shapped arrays are called conformable arrays.
-
-The transformation is based on the observation that Cartesian array processing involves
-identical processing to be performed on different regions of the Cartesian array.
-
-The library will (eventually) both geometrically decompose into MPI tasks and across SIMD lanes.
-
-Data parallel array operations can then be specified with a SINGLE data parallel paradigm, but
-optimally use MPI, OpenMP and SIMD parallelism under the hood. This is a significant simplification
-for most programmers.
-
-The layout transformations are parametrised by the SIMD vector length. This adapts according to the architecture.
-Presently SSE2 (128 bit) AVX, AVX2 (256 bit) and IMCI and AVX512 (512 bit) targets are supported.
-
-These are presented as 
-
-  vRealF, vRealD, vComplexF, vComplexD 
-
-internal vector data types. These may be useful in themselves for other programmers.
-The corresponding scalar types are named
-
-  RealF, RealD, ComplexF, ComplexD
-
-MPI parallelism is UNIMPLEMENTED and for now only OpenMP and SIMD parallelism is present in the library.
-
-   You can give `configure' initial values for configuration parameters
-by setting variables in the command line or in the environment.  Here
-is are examples:
-
-     ./configure CXX=clang++ CXXFLAGS="-std=c++11 -O3 -msse4" --enable-simd=SSE4
-
-     ./configure CXX=clang++ CXXFLAGS="-std=c++11 -O3 -mavx" --enable-simd=AVX1
-
-     ./configure CXX=clang++ CXXFLAGS="-std=c++11 -O3 -mavx2" --enable-simd=AVX2
-
-     ./configure CXX=icpc CXXFLAGS="-std=c++11 -O3 -mmic" --enable-simd=AVX512 --host=none
-     
-     

README

@@ -0,0 +1 @@
+README.md

README.md

@@ -1,26 +1,14 @@
-# Grid
-<table>
-<tr>
-    <td>Last stable release</td>
-    <td><a href="https://travis-ci.org/paboyle/Grid">
-    <img src="https://travis-ci.org/paboyle/Grid.svg?branch=master"></a>
-    </td>
-</tr>
-<tr>
-    <td>Development branch</td>
-    <td><a href="https://travis-ci.org/paboyle/Grid">
-    <img src="https://travis-ci.org/paboyle/Grid.svg?branch=develop"></a>
-    </td>
-</tr>
-</table>
+# Grid [![Teamcity status](http://ci.cliath.ph.ed.ac.uk/app/rest/builds/aggregated/strob:(buildType:(affectedProject(id:Grid)),branch:name:develop)/statusIcon.svg)](http://ci.cliath.ph.ed.ac.uk/project.html?projectId=Grid&tab=projectOverview) [![Travis status](https://travis-ci.org/paboyle/Grid.svg?branch=develop)](https://travis-ci.org/paboyle/Grid)
 
 **Data parallel C++ mathematical object library.**
 
-Please send all pull requests to the `develop` branch.
-
 License: GPL v2.
 
-Last update 2016/08/03.
+Last update June 2017.
+
+_Please do not send pull requests to the `master` branch which is reserved for releases._
+
+
 
 ### Description
 This library provides data parallel C++ container classes with internal memory layout
@@ -29,7 +17,7 @@ are provided, similar to HPF and cmfortran, and user control is given over the m
 array indices to both MPI tasks and SIMD processing elements.
 
 * Identically shaped arrays then be processed with perfect data parallelisation.
-* Such identically shapped arrays are called conformable arrays.
+* Such identically shaped arrays are called conformable arrays.
 
 The transformation is based on the observation that Cartesian array processing involves
 identical processing to be performed on different regions of the Cartesian array.
@@ -42,15 +30,70 @@ optimally use MPI, OpenMP and SIMD parallelism under the hood. This is a signifi
 for most programmers.
 
 The layout transformations are parametrised by the SIMD vector length. This adapts according to the architecture.
-Presently SSE4 (128 bit) AVX, AVX2 (256 bit) and IMCI and AVX512 (512 bit) targets are supported (ARM NEON and BG/Q QPX on the way).
+Presently SSE4, ARM NEON (128 bits) AVX, AVX2, QPX (256 bits), IMCI and AVX512 (512 bits) targets are supported.
 
-These are presented as `vRealF`, `vRealD`, `vComplexF`, and `vComplexD` internal vector data types. These may be useful in themselves for other programmers.
+These are presented as `vRealF`, `vRealD`, `vComplexF`, and `vComplexD` internal vector data types. 
 The corresponding scalar types are named `RealF`, `RealD`, `ComplexF` and `ComplexD`.
 
 MPI, OpenMP, and SIMD parallelism are present in the library.
-Please see https://arxiv.org/abs/1512.03487 for more detail.
+Please see [this paper](https://arxiv.org/abs/1512.03487) for more detail.
 
-### Installation
+
+### Compilers
+
+Intel ICPC v16.0.3 and later
+
+Clang v3.5 and later (need 3.8 and later for OpenMP)
+
+GCC   v4.9.x (recommended)
+
+GCC   v6.3 and later
+
+### Important: 
+
+Some versions of GCC appear to have a bug under high optimisation (-O2, -O3).
+
+The safety of these compiler versions cannot be guaranteed at this time. Follow Issue 100 for details and updates.
+
+GCC   v5.x
+
+GCC   v6.1, v6.2
+
+### Bug report
+
+_To help us tracking and solving more efficiently issues with Grid, please report problems using the issue system of GitHub rather than sending emails to Grid developers._
+
+When you file an issue, please go though the following checklist:
+
+1. Check that the code is pointing to the `HEAD` of `develop` or any commit in `master` which is tagged with a version number. 
+2. Give a description of the target platform (CPU, network, compiler). Please give the full CPU part description, using for example `cat /proc/cpuinfo | grep 'model name' | uniq` (Linux) or `sysctl machdep.cpu.brand_string` (macOS) and the full output the `--version` option of your compiler.
+3. Give the exact `configure` command used.
+4. Attach `config.log`.
+5. Attach `grid.config.summary`.
+6. Attach the output of `make V=1`.
+7. Describe the issue and any previous attempt to solve it. If relevant, show how to reproduce the issue using a minimal working example.
+
+### Required libraries
+Grid requires:
+
+[GMP](https://gmplib.org/), 
+
+[MPFR](http://www.mpfr.org/) 
+
+Bootstrapping grid downloads and uses for internal dense matrix (non-QCD operations) the Eigen library.
+
+Grid optionally uses:
+
+[HDF5](https://support.hdfgroup.org/HDF5/)  
+
+[LIME](http://usqcd-software.github.io/c-lime/) for ILDG and SciDAC file format support. 
+
+[FFTW](http://www.fftw.org) either generic version or via the Intel MKL library.
+
+LAPACK either generic version or Intel MKL library.
+
+
+### Quick start
 First, start by cloning the repository:
 
 ``` bash
@@ -71,20 +114,18 @@ mkdir build; cd build
 ../configure --enable-precision=double --enable-simd=AVX --enable-comms=mpi-auto --prefix=<path>
 ```
 
-where `--enable-precision=` set the default precision (`single` or `double`),
-`--enable-simd=` set the SIMD type (see possible values below), `--enable-
-comms=` set the protocol used for communications (`none`, `mpi`, `mpi-auto` or
-`shmem`), and `<path>` should be replaced by the prefix path where you want to
-install Grid. The `mpi-auto` communication option set `configure` to determine
-automatically how to link to MPI. Other options are available, use `configure
+where `--enable-precision=` set the default precision,
+`--enable-simd=` set the SIMD type, `--enable-
+comms=`, and `<path>` should be replaced by the prefix path where you want to
+install Grid. Other options are detailed in the next section, you can also use `configure
 --help` to display them. Like with any other program using GNU autotool, the
 `CXX`, `CXXFLAGS`, `LDFLAGS`, ... environment variables can be modified to
 customise the build.
 
-Finally, you can build and install Grid:
+Finally, you can build, check, and install Grid:
 
 ``` bash
-make; make install
+make; make check; make install
 ```
 
 To minimise the build time, only the tests at the root of the `tests` directory are built by default. If you want to build tests in the sub-directory `<subdir>` you can execute:
@@ -92,25 +133,274 @@ To minimise the build time, only the tests at the root of the `tests` directory
 ``` bash
 make -C tests/<subdir> tests
 ```
+If you want to build all the tests at once just use `make tests`.
+
+### Build configuration options
+
+- `--prefix=<path>`: installation prefix for Grid.
+- `--with-gmp=<path>`: look for GMP in the UNIX prefix `<path>`
+- `--with-mpfr=<path>`: look for MPFR in the UNIX prefix `<path>`
+- `--with-fftw=<path>`: look for FFTW in the UNIX prefix `<path>`
+- `--enable-lapack[=<path>]`: enable LAPACK support in Lanczos eigensolver. A UNIX prefix containing the library can be specified (optional).
+- `--enable-mkl[=<path>]`: use Intel MKL for FFT (and LAPACK if enabled) routines. A UNIX prefix containing the library can be specified (optional).
+- `--enable-numa`: enable NUMA first touch optimisation
+- `--enable-simd=<code>`: setup Grid for the SIMD target `<code>` (default: `GEN`). A list of possible SIMD targets is detailed in a section below.
+- `--enable-gen-simd-width=<size>`: select the size (in bytes) of the generic SIMD vector type (default: 32 bytes).
+- `--enable-precision={single|double}`: set the default precision (default: `double`).
+- `--enable-precision=<comm>`: Use `<comm>` for message passing (default: `none`). A list of possible SIMD targets is detailed in a section below.
+- `--enable-rng={sitmo|ranlux48|mt19937}`: choose the RNG (default: `sitmo `).
+- `--disable-timers`: disable system dependent high-resolution timers.
+- `--enable-chroma`: enable Chroma regression tests.
+- `--enable-doxygen-doc`: enable the Doxygen documentation generation (build with `make doxygen-doc`)
+
+### Possible communication interfaces
+
+The following options can be use with the `--enable-comms=` option to target different communication interfaces:
+
+| `<comm>`       | Description                                                   |
+| -------------- | ------------------------------------------------------------- |
+| `none`         | no communications                                             |
+| `mpi[-auto]`   | MPI communications                                            |
+| `mpi3[-auto]`  | MPI communications using MPI 3 shared memory                  |
+| `shmem `       | Cray SHMEM communications                                     |
+
+For the MPI interfaces the optional `-auto` suffix instructs the `configure` scripts to determine all the necessary compilation and linking flags. This is done by extracting the informations from the MPI wrapper specified in the environment variable `MPICXX` (if not specified `configure` will scan though a list of default names). The `-auto` suffix is not supported by the Cray environment wrapper scripts. Use the standard versions instead.  
 
 ### Possible SIMD types
 
 The following options can be use with the `--enable-simd=` option to target different SIMD instruction sets:
 
-| String      | Description                            |
+| `<code>`    | Description                            |
 | ----------- | -------------------------------------- |
 | `GEN`       | generic portable vector code           |
 | `SSE4`      | SSE 4.2 (128 bit)                      |
 | `AVX`       | AVX (256 bit)                          |
-| `AVXFMA4`   | AVX (256 bit) + FMA                    |
+| `AVXFMA`    | AVX (256 bit) + FMA                    |
+| `AVXFMA4`   | AVX (256 bit) + FMA4                   |
 | `AVX2`      | AVX 2 (256 bit)                        |
 | `AVX512`    | AVX 512 bit                            |
-| `AVX512MIC` | AVX 512 bit for Intel MIC architecture |
-| `ICMI`      | Intel ICMI instructions (512 bit)      |
+| `NEONv8`    | [ARM NEON](http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.den0024a/ch07s03.html) (128 bit)                     |
+| `QPX`       | IBM QPX (256 bit)                      |
 
 Alternatively, some CPU codenames can be directly used:
 
-| String      | Description                            |
+| `<code>`    | Description                            |
 | ----------- | -------------------------------------- |
-| `KNC`       | [Intel Knights Corner](http://ark.intel.com/products/codename/57721/Knights-Corner) |
-| `KNL`       | [Intel Knights Landing](http://ark.intel.com/products/codename/48999/Knights-Landing) |
+| `KNL`       | [Intel Xeon Phi codename Knights Landing](http://ark.intel.com/products/codename/48999/Knights-Landing) |
+| `BGQ`       | Blue Gene/Q                            |
+
+#### Notes:
+- We currently support AVX512 only for the Intel compiler. Support for GCC and clang will appear in future versions of Grid when the AVX512 support within GCC and clang will be more advanced.
+- For BG/Q only [bgclang](http://trac.alcf.anl.gov/projects/llvm-bgq) is supported. We do not presently plan to support more compilers for this platform.
+- BG/Q performances are currently rather poor. This is being investigated for future versions.
+- The vector size for the `GEN` target can be specified with the `configure` script option `--enable-gen-simd-width`.
+
+### Build setup for Intel Knights Landing platform
+
+The following configuration is recommended for the Intel Knights Landing platform:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=KNL        \
+             --enable-comms=mpi-auto  \
+             --enable-mkl             \
+             CXX=icpc MPICXX=mpiicpc
+```
+The MKL flag enables use of BLAS and FFTW from the Intel Math Kernels Library.
+
+If you are working on a Cray machine that does not use the `mpiicpc` wrapper, please use:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=KNL        \
+             --enable-comms=mpi       \
+             --enable-mkl             \
+             CXX=CC CC=cc
+```
+
+If gmp and mpfr are NOT in standard places (/usr/) these flags may be needed:
+``` bash
+               --with-gmp=<path>        \
+               --with-mpfr=<path>       \
+```
+where `<path>` is the UNIX prefix where GMP and MPFR are installed. 
+
+Knight's Landing with Intel Omnipath adapters with two adapters per node 
+presently performs better with use of more than one rank per node, using shared memory 
+for interior communication. This is the mpi3 communications implementation. 
+We recommend four ranks per node for best performance, but optimum is local volume dependent.
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=KNL        \
+             --enable-comms=mpi3-auto \
+             --enable-mkl             \
+             CC=icpc MPICXX=mpiicpc 
+```
+
+### Build setup for Intel Haswell Xeon platform
+
+The following configuration is recommended for the Intel Haswell platform:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX2       \
+             --enable-comms=mpi3-auto \
+             --enable-mkl             \
+             CXX=icpc MPICXX=mpiicpc
+```
+The MKL flag enables use of BLAS and FFTW from the Intel Math Kernels Library.
+
+If gmp and mpfr are NOT in standard places (/usr/) these flags may be needed:
+``` bash
+               --with-gmp=<path>        \
+               --with-mpfr=<path>       \
+```
+where `<path>` is the UNIX prefix where GMP and MPFR are installed. 
+
+If you are working on a Cray machine that does not use the `mpiicpc` wrapper, please use:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX2       \
+             --enable-comms=mpi3      \
+             --enable-mkl             \
+             CXX=CC CC=cc
+```
+Since Dual socket nodes are commonplace, we recommend MPI-3 as the default with the use of 
+one rank per socket. If using the Intel MPI library, threads should be pinned to NUMA domains using
+```
+        export I_MPI_PIN=1
+```
+This is the default.
+
+### Build setup for Intel Skylake Xeon platform
+
+The following configuration is recommended for the Intel Skylake platform:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX512     \
+             --enable-comms=mpi3      \
+             --enable-mkl             \
+             CXX=mpiicpc
+```
+The MKL flag enables use of BLAS and FFTW from the Intel Math Kernels Library.
+
+If gmp and mpfr are NOT in standard places (/usr/) these flags may be needed:
+``` bash
+               --with-gmp=<path>        \
+               --with-mpfr=<path>       \
+```
+where `<path>` is the UNIX prefix where GMP and MPFR are installed. 
+
+If you are working on a Cray machine that does not use the `mpiicpc` wrapper, please use:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX512     \
+             --enable-comms=mpi3      \
+             --enable-mkl             \
+             CXX=CC CC=cc
+```
+Since Dual socket nodes are commonplace, we recommend MPI-3 as the default with the use of 
+one rank per socket. If using the Intel MPI library, threads should be pinned to NUMA domains using
+``` 
+        export I_MPI_PIN=1
+```
+This is the default. 
+
+#### Expected Skylake Gold 6148 dual socket (single prec, single node 20+20 cores) performance using NUMA MPI mapping): 
+
+mpirun -n 2 benchmarks/Benchmark_dwf --grid 16.16.16.16 --mpi 2.1.1.1 --cacheblocking 2.2.2.2 --dslash-asm --shm 1024 --threads 18 
+
+TBA
+
+
+### Build setup for AMD EPYC / RYZEN
+
+The AMD EPYC is a multichip module comprising 32 cores spread over four distinct chips each with 8 cores.
+So, even with a single socket node there is a quad-chip module. Dual socket nodes with 64 cores total
+are common. Each chip within the module exposes a separate NUMA domain.
+There are four NUMA domains per socket and we recommend one MPI rank per NUMA domain.
+MPI-3 is recommended with the use of four ranks per socket,
+and 8 threads per rank. 
+
+The following configuration is recommended for the AMD EPYC platform.
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX2       \
+             --enable-comms=mpi3 \
+             CXX=mpicxx 
+```
+
+If gmp and mpfr are NOT in standard places (/usr/) these flags may be needed:
+``` bash
+               --with-gmp=<path>        \
+               --with-mpfr=<path>       \
+```
+where `<path>` is the UNIX prefix where GMP and MPFR are installed. 
+
+Using MPICH and g++ v4.9.2, best performance can be obtained using explicit GOMP_CPU_AFFINITY flags for each MPI rank.
+This can be done by invoking MPI on a wrapper script omp_bind.sh to handle this. 
+
+It is recommended to run 8 MPI ranks on a single dual socket AMD EPYC, with 8 threads per rank using MPI3 and
+shared memory to communicate within this node:
+
+mpirun -np 8 ./omp_bind.sh ./Benchmark_dwf --mpi 2.2.2.1 --dslash-unroll --threads 8 --grid 16.16.16.16 --cacheblocking 4.4.4.4 
+
+Where omp_bind.sh does the following:
+```
+#!/bin/bash
+
+numanode=` expr $PMI_RANK % 8 `
+basecore=`expr $numanode \* 16`
+core0=`expr $basecore + 0 `
+core1=`expr $basecore + 2 `
+core2=`expr $basecore + 4 `
+core3=`expr $basecore + 6 `
+core4=`expr $basecore + 8 `
+core5=`expr $basecore + 10 `
+core6=`expr $basecore + 12 `
+core7=`expr $basecore + 14 `
+
+export GOMP_CPU_AFFINITY="$core0 $core1 $core2 $core3 $core4 $core5 $core6 $core7"
+echo GOMP_CUP_AFFINITY $GOMP_CPU_AFFINITY
+
+$@
+```
+
+Performance:
+
+#### Expected AMD EPYC 7601 dual socket (single prec, single node 32+32 cores) performance using NUMA MPI mapping): 
+
+mpirun  -np 8 ./omp_bind.sh ./Benchmark_dwf --threads 8 --mpi 2.2.2.1 --dslash-unroll --grid 16.16.16.16 --cacheblocking 4.4.4.4
+
+TBA
+
+### Build setup for BlueGene/Q
+
+To be written...
+
+### Build setup for ARM Neon
+
+To be written...
+
+### Build setup for laptops, other compilers, non-cluster builds
+
+Many versions of g++ and clang++ work with Grid, and involve merely replacing CXX (and MPICXX),
+and omit the enable-mkl flag. 
+
+Single node builds are enabled with 
+```
+            --enable-comms=none
+```
+
+FFTW support that is not in the default search path may then enabled with
+```
+    --with-fftw=<installpath>
+```
+
+BLAS will not be compiled in by default, and Lanczos will default to Eigen diagonalisation.
+

TODO

@@ -1,6 +1,35 @@
 TODO:
 ---------------
 
+Large item work list:
+
+1)- BG/Q port and check
+2)- Christoph's local basis expansion Lanczos
+3)- Precision conversion and sort out localConvert      <-- partial
+
+  - Consistent linear solver flop count/rate -- PARTIAL, time but no flop/s yet
+4)- Physical propagator interface
+5)- Conserved currents
+6)- Multigrid Wilson and DWF, compare to other Multigrid implementations
+7)- HDCR resume
+
+Recent DONE 
+
+-- MultiRHS with spread out extra dim -- Go through filesystem with SciDAC I/O.  <--- DONE
+-- Lanczos Remove DenseVector, DenseMatrix; Use Eigen instead. <-- DONE
+-- GaugeFix into central location                      <-- DONE
+-- Scidac and Ildg metadata handling                   <-- DONE
+-- Binary I/O MPI2 IO                                  <-- DONE
+-- Binary I/O speed up & x-strips                      <-- DONE
+-- Cut down the exterior overhead                      <-- DONE
+-- Interior legs from SHM comms                        <-- DONE
+-- Half-precision comms                                <-- DONE
+-- Merge high precision reduction into develop         <-- DONE
+-- BlockCG, BCGrQ                                      <-- DONE
+-- multiRHS DWF; benchmark on Cori/BNL for comms elimination <-- DONE
+   -- slice* linalg routines for multiRHS, BlockCG    
+
+-----
 * Forces; the UdSdU  term in gauge force term is half of what I think it should
   be. This is a consequence of taking ONLY the first term in:
 
@@ -21,16 +50,8 @@ TODO:
   This means we must double the force in the Test_xxx_force routines, and is the origin of the factor of two.
   This 2x is applied by hand in the fermion routines and in the Test_rect_force routine.
 
-
-Policies:
-
-* Link smearing/boundary conds; Policy class based implementation ; framework more in place
-
 * Support different boundary conditions (finite temp, chem. potential ... )
 
-* Support different fermion representations? 
-  - contained entirely within the integrator presently
-
 - Sign of force term.
 
 - Reversibility test.
@@ -41,11 +62,6 @@ Policies:
 
 - Audit oIndex usage for cb behaviour
 
-- Rectangle gauge actions.
-  Iwasaki,
-  Symanzik,
-  ... etc...
-
 - Prepare multigrid for HMC. - Alternate setup schemes.
 
 - Support for ILDG --- ugly, not done
@@ -55,9 +71,11 @@ Policies:
 - FFTnD ?
 
 - Gparity; hand opt use template specialisation elegance to enable the optimised paths ?
+
 - Gparity force term; Gparity (R)HMC.
-- Random number state save restore
+
 - Mobius implementation clean up to rmove #if 0 stale code sequences
+
 - CG -- profile carefully, kernel fusion, whole CG performance measurements.
 
 ================================================================
@@ -90,6 +108,7 @@ Insert/Extract
 Not sure of status of this -- reverify. Things are working nicely now though.
 
 * Make the Tensor types and Complex etc... play more nicely.
+
   - TensorRemove is a hack, come up with a long term rationalised approach to Complex vs. Scalar<Scalar<Scalar<Complex > > >
     QDP forces use of "toDouble" to get back to non tensor scalar. This role is presently taken TensorRemove, but I
     want to introduce a syntax that does not require this.
@@ -112,6 +131,8 @@ Not sure of status of this -- reverify. Things are working nicely now though.
 RECENT
 ---------------
 
+  - Support different fermion representations? -- DONE
+  - contained entirely within the integrator presently
   - Clean up HMC                                                             -- DONE
   - LorentzScalar<GaugeField> gets Gauge link type (cleaner).                -- DONE
   - Simplified the integrators a bit.                                        -- DONE
@@ -123,6 +144,26 @@ RECENT
   - Parallel io improvements                                  -- DONE
   - Plaquette and link trace checks into nersc reader from the Grid_nersc_io.cc test. -- DONE
 
+
+DONE:
+- MultiArray -- MultiRHS done
+- ConjugateGradientMultiShift -- DONE
+- MCR                         -- DONE
+- Remez -- Mike or Boost?     -- DONE
+- Proto (ET)                  -- DONE
+- uBlas                       -- DONE ; Eigen
+- Potentially Useful Boost libraries -- DONE ; Eigen
+- Aligned allocator; memory pool -- DONE
+- Multiprecision              -- DONE
+- Serialization               -- DONE
+- Regex -- Not needed
+- Tokenize -- Why?
+
+- Random number state save restore -- DONE
+- Rectangle gauge actions. -- DONE
+  Iwasaki,
+  Symanzik,
+  ... etc...
 Done: Cayley, Partial , ContFrac force terms.
 
 DONE
@@ -207,6 +248,7 @@ Done
 FUNCTIONALITY: it pleases me to keep track of things I have done (keeps me arguably sane)
 ======================================================================================================
 
+* Link smearing/boundary conds; Policy class based implementation ; framework more in place -- DONE
 * Command line args for geometry, simd, etc. layout. Is it necessary to have -- DONE
   user pass these? Is this a QCD specific?
 

VERSION

@@ -1,4 +1,5 @@
-Version : 0.5.0
+Version : 0.7.0
 
-- AVX512, AVX2, AVX, SSE good
-- Clang 3.5 and above, ICPC v16 and above, GCC 4.9 and above
+- Clang 3.5 and above, ICPC v16 and above, GCC 6.3 and above recommended
+- MPI and MPI3 comms optimisations for KNL and OPA finished
+- Half precision comms

benchmarks/Benchmark_ITT.cc

@@ -0,0 +1,797 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./benchmarks/Benchmark_memory_bandwidth.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+typedef WilsonFermion5D<DomainWallVec5dImplR> WilsonFermion5DR;
+typedef WilsonFermion5D<DomainWallVec5dImplF> WilsonFermion5DF;
+typedef WilsonFermion5D<DomainWallVec5dImplD> WilsonFermion5DD;
+
+
+std::vector<int> L_list;
+std::vector<int> Ls_list;
+std::vector<double> mflop_list;
+
+double mflop_ref;
+double mflop_ref_err;
+
+int NN_global;
+
+struct time_statistics{
+  double mean;
+  double err;
+  double min;
+  double max;
+
+  void statistics(std::vector<double> v){
+      double sum = std::accumulate(v.begin(), v.end(), 0.0);
+      mean = sum / v.size();
+
+      std::vector<double> diff(v.size());
+      std::transform(v.begin(), v.end(), diff.begin(), [=](double x) { return x - mean; });
+      double sq_sum = std::inner_product(diff.begin(), diff.end(), diff.begin(), 0.0);
+      err = std::sqrt(sq_sum / (v.size()*(v.size() - 1)));
+
+      auto result = std::minmax_element(v.begin(), v.end());
+      min = *result.first;
+      max = *result.second;
+}
+};
+
+void comms_header(){
+  std::cout <<GridLogMessage << " L  "<<"\t"<<" Ls  "<<"\t"
+            <<std::setw(11)<<"bytes"<<"MB/s uni (err/min/max)"<<"\t\t"<<"MB/s bidi (err/min/max)"<<std::endl;
+};
+
+Gamma::Algebra Gmu [] = {
+  Gamma::Algebra::GammaX,
+  Gamma::Algebra::GammaY,
+  Gamma::Algebra::GammaZ,
+  Gamma::Algebra::GammaT
+};
+struct controls {
+  int Opt;
+  int CommsOverlap;
+  Grid::CartesianCommunicator::CommunicatorPolicy_t CommsAsynch;
+  //  int HugePages;
+};
+
+class Benchmark {
+public:
+  static void Decomposition (void ) {
+
+    int threads = GridThread::GetThreads();
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << "= Grid is setup to use "<<threads<<" threads"<<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage<<"Grid Default Decomposition patterns\n";
+    std::cout<<GridLogMessage<<"\tOpenMP threads : "<<GridThread::GetThreads()<<std::endl;
+    std::cout<<GridLogMessage<<"\tMPI tasks      : "<<GridCmdVectorIntToString(GridDefaultMpi())<<std::endl;
+    std::cout<<GridLogMessage<<"\tvReal          : "<<sizeof(vReal )*8    <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vReal::Nsimd()))<<std::endl;
+    std::cout<<GridLogMessage<<"\tvRealF         : "<<sizeof(vRealF)*8    <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vRealF::Nsimd()))<<std::endl;
+    std::cout<<GridLogMessage<<"\tvRealD         : "<<sizeof(vRealD)*8    <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vRealD::Nsimd()))<<std::endl;
+    std::cout<<GridLogMessage<<"\tvComplex       : "<<sizeof(vComplex )*8 <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vComplex::Nsimd()))<<std::endl;
+    std::cout<<GridLogMessage<<"\tvComplexF      : "<<sizeof(vComplexF)*8 <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vComplexF::Nsimd()))<<std::endl;
+    std::cout<<GridLogMessage<<"\tvComplexD      : "<<sizeof(vComplexD)*8 <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vComplexD::Nsimd()))<<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+  }
+
+  static void Comms(void)
+  {
+    int Nloop=200;
+    int nmu=0;
+    int maxlat=32;
+
+    std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
+    std::vector<int> mpi_layout  = GridDefaultMpi();
+
+    for(int mu=0;mu<Nd;mu++) if (mpi_layout[mu]>1) nmu++;
+
+    std::vector<double> t_time(Nloop);
+    time_statistics timestat;
+
+    std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << "= Benchmarking threaded STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
+    std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+    comms_header();
+
+    for(int lat=4;lat<=maxlat;lat+=4){
+      for(int Ls=8;Ls<=8;Ls*=2){
+
+	std::vector<int> latt_size  ({lat*mpi_layout[0],
+	      lat*mpi_layout[1],
+	      lat*mpi_layout[2],
+	      lat*mpi_layout[3]});
+
+	GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+	RealD Nrank = Grid._Nprocessors;
+	RealD Nnode = Grid.NodeCount();
+	RealD ppn = Nrank/Nnode;
+
+	std::vector<HalfSpinColourVectorD *> xbuf(8);
+	std::vector<HalfSpinColourVectorD *> rbuf(8);
+	Grid.ShmBufferFreeAll();
+	for(int d=0;d<8;d++){
+	  xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	  rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	  bzero((void *)xbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	  bzero((void *)rbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	}
+
+	int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
+	int ncomm;
+	double dbytes;
+	std::vector<double> times(Nloop);
+	for(int i=0;i<Nloop;i++){
+
+	  double start=usecond();
+
+	  dbytes=0;
+	  ncomm=0;
+
+	  parallel_for(int dir=0;dir<8;dir++){
+
+	    double tbytes;
+	    int mu =dir % 4;
+
+	    if (mpi_layout[mu]>1 ) {
+	        
+	      int xmit_to_rank;
+	      int recv_from_rank;
+	      if ( dir == mu ) { 
+		int comm_proc=1;
+		Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	      } else { 
+		int comm_proc = mpi_layout[mu]-1;
+		Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	      }
+	      tbytes= Grid.StencilSendToRecvFrom((void *)&xbuf[dir][0], xmit_to_rank,
+						 (void *)&rbuf[dir][0], recv_from_rank,
+						 bytes,dir);
+	  
+#ifdef GRID_OMP
+#pragma omp atomic
+#endif
+	      ncomm++;
+
+#ifdef GRID_OMP
+#pragma omp atomic
+#endif
+	      dbytes+=tbytes;
+	    }
+	  }
+	  Grid.Barrier();
+	  double stop=usecond();
+	  t_time[i] = stop-start; // microseconds
+	}
+
+	timestat.statistics(t_time);
+	//	for(int i=0;i<t_time.size();i++){
+	//	  std::cout << i<<" "<<t_time[i]<<std::endl;
+	//	}
+
+	dbytes=dbytes*ppn;
+	double xbytes    = dbytes*0.5;
+	double rbytes    = dbytes*0.5;
+	double bidibytes = dbytes;
+
+	std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
+		 <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
+		 <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
+		 <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
+		 << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
+		 << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
+
+ 
+	
+	    }
+    }    
+
+    return;
+  }
+
+  static void Memory(void)
+  {
+    const int Nvec=8;
+    typedef Lattice< iVector< vReal,Nvec> > LatticeVec;
+    typedef iVector<vReal,Nvec> Vec;
+
+    std::vector<int> simd_layout = GridDefaultSimd(Nd,vReal::Nsimd());
+    std::vector<int> mpi_layout  = GridDefaultMpi();
+
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << "= Benchmarking a*x + y bandwidth"<<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s"<<"\t\t"<<"Gflop/s"<<"\t\t seconds"<< "\t\tGB/s / node"<<std::endl;
+    std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
+  
+    uint64_t NP;
+    uint64_t NN;
+
+
+  uint64_t lmax=48;
+#define NLOOP (100*lmax*lmax*lmax*lmax/lat/lat/lat/lat)
+
+    GridSerialRNG          sRNG;      sRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+    for(int lat=8;lat<=lmax;lat+=4){
+
+      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+
+      NP= Grid.RankCount();
+      NN =Grid.NodeCount();
+
+      Vec rn ; random(sRNG,rn);
+
+      LatticeVec z(&Grid); z=rn;
+      LatticeVec x(&Grid); x=rn;
+      LatticeVec y(&Grid); y=rn;
+      double a=2.0;
+
+      uint64_t Nloop=NLOOP;
+
+      double start=usecond();
+      for(int i=0;i<Nloop;i++){
+	z=a*x-y;
+        x._odata[0]=z._odata[0]; // force serial dependency to prevent optimise away
+        y._odata[4]=z._odata[4];
+      }
+      double stop=usecond();
+      double time = (stop-start)/Nloop*1000;
+     
+      double flops=vol*Nvec*2;// mul,add
+      double bytes=3.0*vol*Nvec*sizeof(Real);
+      std::cout<<GridLogMessage<<std::setprecision(3) 
+	       << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t"<<flops/time<<"\t\t"<<(stop-start)/1000./1000.
+	       << "\t\t"<< bytes/time/NN <<std::endl;
+
+    }
+  };
+
+  static double DWF5(int Ls,int L)
+  {
+    RealD mass=0.1;
+    RealD M5  =1.8;
+
+    double mflops;
+    double mflops_best = 0;
+    double mflops_worst= 0;
+    std::vector<double> mflops_all;
+
+    ///////////////////////////////////////////////////////
+    // Set/Get the layout & grid size
+    ///////////////////////////////////////////////////////
+    int threads = GridThread::GetThreads();
+    std::vector<int> mpi = GridDefaultMpi(); assert(mpi.size()==4);
+    std::vector<int> local({L,L,L,L});
+
+    GridCartesian         * TmpGrid   = SpaceTimeGrid::makeFourDimGrid(std::vector<int>({64,64,64,64}), 
+								       GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+    uint64_t NP = TmpGrid->RankCount();
+    uint64_t NN = TmpGrid->NodeCount();
+    NN_global=NN;
+    uint64_t SHM=NP/NN;
+
+    std::vector<int> internal;
+    if      ( SHM == 1 )   internal = std::vector<int>({1,1,1,1});
+    else if ( SHM == 2 )   internal = std::vector<int>({2,1,1,1});
+    else if ( SHM == 4 )   internal = std::vector<int>({2,2,1,1});
+    else if ( SHM == 8 )   internal = std::vector<int>({2,2,2,1});
+    else assert(0);
+
+    std::vector<int> nodes({mpi[0]/internal[0],mpi[1]/internal[1],mpi[2]/internal[2],mpi[3]/internal[3]});
+    std::vector<int> latt4({local[0]*nodes[0],local[1]*nodes[1],local[2]*nodes[2],local[3]*nodes[3]});
+
+    ///////// Welcome message ////////////
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << "Benchmark DWF Ls vec on "<<L<<"^4 local volume "<<std::endl;
+    std::cout<<GridLogMessage << "* Global volume  : "<<GridCmdVectorIntToString(latt4)<<std::endl;
+    std::cout<<GridLogMessage << "* Ls             : "<<Ls<<std::endl;
+    std::cout<<GridLogMessage << "* MPI ranks      : "<<GridCmdVectorIntToString(mpi)<<std::endl;
+    std::cout<<GridLogMessage << "* Intranode      : "<<GridCmdVectorIntToString(internal)<<std::endl;
+    std::cout<<GridLogMessage << "* nodes          : "<<GridCmdVectorIntToString(nodes)<<std::endl;
+    std::cout<<GridLogMessage << "* Using "<<threads<<" threads"<<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+    ///////// Lattice Init ////////////
+    GridCartesian         * UGrid    = SpaceTimeGrid::makeFourDimGrid(latt4, GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+    GridRedBlackCartesian * UrbGrid  = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+    GridCartesian         * sUGrid   = SpaceTimeGrid::makeFourDimDWFGrid(latt4,GridDefaultMpi());
+    GridRedBlackCartesian * sUrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(sUGrid);
+    GridCartesian         * sFGrid   = SpaceTimeGrid::makeFiveDimDWFGrid(Ls,UGrid);
+    GridRedBlackCartesian * sFrbGrid = SpaceTimeGrid::makeFiveDimDWFRedBlackGrid(Ls,UGrid);
+
+    ///////// RNG Init ////////////
+    std::vector<int> seeds4({1,2,3,4});
+    std::vector<int> seeds5({5,6,7,8});
+    GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
+    GridParallelRNG          RNG5(sFGrid);  RNG5.SeedFixedIntegers(seeds5);
+    std::cout << GridLogMessage << "Initialised RNGs" << std::endl;
+
+    ///////// Source preparation ////////////
+    LatticeFermion src   (sFGrid); random(RNG5,src);
+    LatticeFermion tmp   (sFGrid);
+
+    RealD N2 = 1.0/::sqrt(norm2(src));
+    src = src*N2;
+    
+    LatticeGaugeField Umu(UGrid);  SU3::HotConfiguration(RNG4,Umu); 
+
+    WilsonFermion5DR sDw(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,M5);
+    LatticeFermion src_e (sFrbGrid);
+    LatticeFermion src_o (sFrbGrid);
+    LatticeFermion r_e   (sFrbGrid);
+    LatticeFermion r_o   (sFrbGrid);
+    LatticeFermion r_eo  (sFGrid);
+    LatticeFermion err   (sFGrid);
+    {
+
+      pickCheckerboard(Even,src_e,src);
+      pickCheckerboard(Odd,src_o,src);
+
+#if defined(AVX512) 
+      const int num_cases = 6;
+      std::string fmt("A/S ; A/O ; U/S ; U/O ; G/S ; G/O ");
+#else
+      const int num_cases = 4;
+      std::string fmt("U/S ; U/O ; G/S ; G/O ");
+#endif
+      controls Cases [] = {
+#ifdef AVX512
+	{ QCD::WilsonKernelsStatic::OptInlineAsm , QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptInlineAsm , QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
+#endif
+	{ QCD::WilsonKernelsStatic::OptHandUnroll, QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptHandUnroll, QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptGeneric   , QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptGeneric   , QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  }
+      }; 
+
+      for(int c=0;c<num_cases;c++) {
+
+	QCD::WilsonKernelsStatic::Comms = Cases[c].CommsOverlap;
+	QCD::WilsonKernelsStatic::Opt   = Cases[c].Opt;
+	CartesianCommunicator::SetCommunicatorPolicy(Cases[c].CommsAsynch);
+
+	std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+	if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
+	if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
+	if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
+	if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
+	if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
+	if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
+	if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
+	std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+	int nwarm = 100;
+	uint64_t ncall = 1000;
+
+	double t0=usecond();
+	sFGrid->Barrier();
+	for(int i=0;i<nwarm;i++){
+	  sDw.DhopEO(src_o,r_e,DaggerNo);
+	}
+	sFGrid->Barrier();
+	double t1=usecond();
+
+	sDw.ZeroCounters();
+	time_statistics timestat;
+	std::vector<double> t_time(ncall);
+	for(uint64_t i=0;i<ncall;i++){
+	  t0=usecond();
+	  sDw.DhopEO(src_o,r_e,DaggerNo);
+	  t1=usecond();
+	  t_time[i] = t1-t0;
+	}
+	sFGrid->Barrier();
+	
+	double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
+	double flops=(1344.0*volume)/2;
+	double mf_hi, mf_lo, mf_err;
+
+	timestat.statistics(t_time);
+	mf_hi = flops/timestat.min;
+	mf_lo = flops/timestat.max;
+	mf_err= flops/timestat.min * timestat.err/timestat.mean;
+
+	mflops = flops/timestat.mean;
+	mflops_all.push_back(mflops);
+	if ( mflops_best == 0   ) mflops_best = mflops;
+	if ( mflops_worst== 0   ) mflops_worst= mflops;
+	if ( mflops>mflops_best ) mflops_best = mflops;
+	if ( mflops<mflops_worst) mflops_worst= mflops;
+
+	std::cout<<GridLogMessage << std::fixed << std::setprecision(1)<<"sDeo mflop/s =   "<< mflops << " ("<<mf_err<<") " << mf_lo<<"-"<<mf_hi <<std::endl;
+	std::cout<<GridLogMessage << std::fixed << std::setprecision(1)<<"sDeo mflop/s per rank   "<< mflops/NP<<std::endl;
+	std::cout<<GridLogMessage << std::fixed << std::setprecision(1)<<"sDeo mflop/s per node   "<< mflops/NN<<std::endl;
+
+	sDw.Report();
+
+      }
+      double robust = mflops_worst/mflops_best;;
+      std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+      std::cout<<GridLogMessage << L<<"^4 x "<<Ls<< " sDeo Best  mflop/s        =   "<< mflops_best << " ; " << mflops_best/NN<<" per node " <<std::endl;
+      std::cout<<GridLogMessage << L<<"^4 x "<<Ls<< " sDeo Worst mflop/s        =   "<< mflops_worst<< " ; " << mflops_worst/NN<<" per node " <<std::endl;
+
+      std::cout<<GridLogMessage <<std::setprecision(3)<< L<<"^4 x "<<Ls<< " Performance Robustness   =   "<< robust <<std::endl;
+      std::cout<<GridLogMessage <<fmt << std::endl;
+      std::cout<<GridLogMessage;
+
+      for(int i=0;i<mflops_all.size();i++){
+	std::cout<<mflops_all[i]/NN<<" ; " ;
+      }
+      std::cout<<std::endl;
+      std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+    }
+    return mflops_best;
+  }
+
+  static double DWF(int Ls,int L, double & robust)
+  {
+    RealD mass=0.1;
+    RealD M5  =1.8;
+
+    double mflops;
+    double mflops_best = 0;
+    double mflops_worst= 0;
+    std::vector<double> mflops_all;
+
+    ///////////////////////////////////////////////////////
+    // Set/Get the layout & grid size
+    ///////////////////////////////////////////////////////
+    int threads = GridThread::GetThreads();
+    std::vector<int> mpi = GridDefaultMpi(); assert(mpi.size()==4);
+    std::vector<int> local({L,L,L,L});
+
+    GridCartesian         * TmpGrid   = SpaceTimeGrid::makeFourDimGrid(std::vector<int>({64,64,64,64}), 
+								       GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+    uint64_t NP = TmpGrid->RankCount();
+    uint64_t NN = TmpGrid->NodeCount();
+    NN_global=NN;
+    uint64_t SHM=NP/NN;
+
+    std::vector<int> internal;
+    if      ( SHM == 1 )   internal = std::vector<int>({1,1,1,1});
+    else if ( SHM == 2 )   internal = std::vector<int>({2,1,1,1});
+    else if ( SHM == 4 )   internal = std::vector<int>({2,2,1,1});
+    else if ( SHM == 8 )   internal = std::vector<int>({2,2,2,1});
+    else assert(0);
+
+    std::vector<int> nodes({mpi[0]/internal[0],mpi[1]/internal[1],mpi[2]/internal[2],mpi[3]/internal[3]});
+    std::vector<int> latt4({local[0]*nodes[0],local[1]*nodes[1],local[2]*nodes[2],local[3]*nodes[3]});
+
+    ///////// Welcome message ////////////
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << "Benchmark DWF on "<<L<<"^4 local volume "<<std::endl;
+    std::cout<<GridLogMessage << "* Global volume  : "<<GridCmdVectorIntToString(latt4)<<std::endl;
+    std::cout<<GridLogMessage << "* Ls             : "<<Ls<<std::endl;
+    std::cout<<GridLogMessage << "* MPI ranks      : "<<GridCmdVectorIntToString(mpi)<<std::endl;
+    std::cout<<GridLogMessage << "* Intranode      : "<<GridCmdVectorIntToString(internal)<<std::endl;
+    std::cout<<GridLogMessage << "* nodes          : "<<GridCmdVectorIntToString(nodes)<<std::endl;
+    std::cout<<GridLogMessage << "* Using "<<threads<<" threads"<<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+
+    ///////// Lattice Init ////////////
+    GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(latt4, GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+    GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+    GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+    GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+
+    
+    ///////// RNG Init ////////////
+    std::vector<int> seeds4({1,2,3,4});
+    std::vector<int> seeds5({5,6,7,8});
+    GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
+    GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
+    std::cout << GridLogMessage << "Initialised RNGs" << std::endl;
+
+    ///////// Source preparation ////////////
+    LatticeFermion src   (FGrid); random(RNG5,src);
+    LatticeFermion ref   (FGrid);
+    LatticeFermion tmp   (FGrid);
+
+    RealD N2 = 1.0/::sqrt(norm2(src));
+    src = src*N2;
+    
+    LatticeGaugeField Umu(UGrid);  SU3::HotConfiguration(RNG4,Umu); 
+
+    DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+
+    ////////////////////////////////////
+    // Naive wilson implementation
+    ////////////////////////////////////
+    {
+      LatticeGaugeField Umu5d(FGrid); 
+      std::vector<LatticeColourMatrix> U(4,FGrid);
+      for(int ss=0;ss<Umu._grid->oSites();ss++){
+	for(int s=0;s<Ls;s++){
+	  Umu5d._odata[Ls*ss+s] = Umu._odata[ss];
+	}
+      }
+      ref = zero;
+      for(int mu=0;mu<Nd;mu++){
+	U[mu] = PeekIndex<LorentzIndex>(Umu5d,mu);
+      }
+      for(int mu=0;mu<Nd;mu++){
+	
+	tmp = U[mu]*Cshift(src,mu+1,1);
+	ref=ref + tmp - Gamma(Gmu[mu])*tmp;
+	
+	tmp =adj(U[mu])*src;
+	tmp =Cshift(tmp,mu+1,-1);
+	ref=ref + tmp + Gamma(Gmu[mu])*tmp;
+      }
+      ref = -0.5*ref;
+    }
+
+    LatticeFermion src_e (FrbGrid);
+    LatticeFermion src_o (FrbGrid);
+    LatticeFermion r_e   (FrbGrid);
+    LatticeFermion r_o   (FrbGrid);
+    LatticeFermion r_eo  (FGrid);
+    LatticeFermion err   (FGrid);
+    {
+
+      pickCheckerboard(Even,src_e,src);
+      pickCheckerboard(Odd,src_o,src);
+
+#if defined(AVX512) 
+      const int num_cases = 6;
+      std::string fmt("A/S ; A/O ; U/S ; U/O ; G/S ; G/O ");
+#else
+      const int num_cases = 4;
+      std::string fmt("U/S ; U/O ; G/S ; G/O ");
+#endif
+      controls Cases [] = {
+#ifdef AVX512
+	{ QCD::WilsonKernelsStatic::OptInlineAsm , QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptInlineAsm , QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
+#endif
+	{ QCD::WilsonKernelsStatic::OptHandUnroll, QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptHandUnroll, QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptGeneric   , QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptGeneric   , QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  }
+      }; 
+
+      for(int c=0;c<num_cases;c++) {
+
+	QCD::WilsonKernelsStatic::Comms = Cases[c].CommsOverlap;
+	QCD::WilsonKernelsStatic::Opt   = Cases[c].Opt;
+	CartesianCommunicator::SetCommunicatorPolicy(Cases[c].CommsAsynch);
+
+	std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+	if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
+	if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
+	if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
+	if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
+	if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
+	if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
+	if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
+	std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+	int nwarm = 200;
+	double t0=usecond();
+	FGrid->Barrier();
+	for(int i=0;i<nwarm;i++){
+	  Dw.DhopEO(src_o,r_e,DaggerNo);
+	}
+	FGrid->Barrier();
+	double t1=usecond();
+	//	uint64_t ncall = (uint64_t) 2.5*1000.0*1000.0*nwarm/(t1-t0);
+	//	if (ncall < 500) ncall = 500;
+	uint64_t ncall = 1000;
+
+	FGrid->Broadcast(0,&ncall,sizeof(ncall));
+
+	//	std::cout << GridLogMessage << " Estimate " << ncall << " calls per second"<<std::endl;
+	Dw.ZeroCounters();
+
+	time_statistics timestat;
+	std::vector<double> t_time(ncall);
+	for(uint64_t i=0;i<ncall;i++){
+	  t0=usecond();
+	  Dw.DhopEO(src_o,r_e,DaggerNo);
+	  t1=usecond();
+	  t_time[i] = t1-t0;
+	}
+	FGrid->Barrier();
+	
+	double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
+	double flops=(1344.0*volume)/2;
+	double mf_hi, mf_lo, mf_err;
+
+	timestat.statistics(t_time);
+	mf_hi = flops/timestat.min;
+	mf_lo = flops/timestat.max;
+	mf_err= flops/timestat.min * timestat.err/timestat.mean;
+
+	mflops = flops/timestat.mean;
+	mflops_all.push_back(mflops);
+	if ( mflops_best == 0   ) mflops_best = mflops;
+	if ( mflops_worst== 0   ) mflops_worst= mflops;
+	if ( mflops>mflops_best ) mflops_best = mflops;
+	if ( mflops<mflops_worst) mflops_worst= mflops;
+
+	std::cout<<GridLogMessage << std::fixed << std::setprecision(1)<<"Deo mflop/s =   "<< mflops << " ("<<mf_err<<") " << mf_lo<<"-"<<mf_hi <<std::endl;
+	std::cout<<GridLogMessage << std::fixed << std::setprecision(1)<<"Deo mflop/s per rank   "<< mflops/NP<<std::endl;
+	std::cout<<GridLogMessage << std::fixed << std::setprecision(1)<<"Deo mflop/s per node   "<< mflops/NN<<std::endl;
+
+	Dw.Report();
+
+	Dw.DhopEO(src_o,r_e,DaggerNo);
+	Dw.DhopOE(src_e,r_o,DaggerNo);
+	setCheckerboard(r_eo,r_o);
+	setCheckerboard(r_eo,r_e);
+	err = r_eo-ref; 
+	std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
+	assert((norm2(err)<1.0e-4));
+
+      }
+      robust = mflops_worst/mflops_best;
+      std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+      std::cout<<GridLogMessage << L<<"^4 x "<<Ls<< " Deo Best  mflop/s        =   "<< mflops_best << " ; " << mflops_best/NN<<" per node " <<std::endl;
+      std::cout<<GridLogMessage << L<<"^4 x "<<Ls<< " Deo Worst mflop/s        =   "<< mflops_worst<< " ; " << mflops_worst/NN<<" per node " <<std::endl;
+      std::cout<<GridLogMessage << std::fixed<<std::setprecision(3)<< L<<"^4 x "<<Ls<< " Performance Robustness   =   "<< robust  <<std::endl;
+      std::cout<<GridLogMessage <<fmt << std::endl;
+      std::cout<<GridLogMessage ;
+
+      for(int i=0;i<mflops_all.size();i++){
+	std::cout<<mflops_all[i]/NN<<" ; " ;
+      }
+      std::cout<<std::endl;
+      std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+    }
+    return mflops_best;
+  }
+
+};
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  CartesianCommunicator::SetCommunicatorPolicy(CartesianCommunicator::CommunicatorPolicySequential);
+#ifdef KNL
+  LebesgueOrder::Block = std::vector<int>({8,2,2,2});
+#else
+  LebesgueOrder::Block = std::vector<int>({2,2,2,2});
+#endif
+  Benchmark::Decomposition();
+
+  int do_memory=1;
+  int do_comms =1;
+  int do_su3   =0;
+  int do_wilson=1;
+  int do_dwf   =1;
+
+  if ( do_su3 ) {
+    // empty for now
+  }
+
+  int sel=2;
+  std::vector<int> L_list({8,12,16,24});
+
+  //int sel=1;
+  //  std::vector<int> L_list({8,12});
+  std::vector<double> robust_list;
+
+  std::vector<double> wilson;
+  std::vector<double> dwf4;
+  std::vector<double> dwf5;
+
+  if ( do_wilson ) {
+    int Ls=1;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << " Wilson dslash 4D vectorised" <<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    for(int l=0;l<L_list.size();l++){
+      double robust;
+      wilson.push_back(Benchmark::DWF(1,L_list[l],robust));
+    }
+  }
+
+  int Ls=16;
+  if ( do_dwf ) {
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << " Domain wall dslash 4D vectorised" <<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    for(int l=0;l<L_list.size();l++){
+      double robust;
+      double result = Benchmark::DWF(Ls,L_list[l],robust) ;
+      dwf4.push_back(result);
+      robust_list.push_back(robust);
+    }
+  }
+
+  if ( do_dwf ) {
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << " Domain wall dslash 4D vectorised" <<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    for(int l=0;l<L_list.size();l++){
+      dwf5.push_back(Benchmark::DWF5(Ls,L_list[l]));
+    }
+
+  }
+
+  if ( do_dwf ) {
+
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << " Summary table Ls="<<Ls <<std::endl;
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "L \t\t Wilson \t DWF4 \t DWF5 " <<std::endl;
+  for(int l=0;l<L_list.size();l++){
+    std::cout<<GridLogMessage << L_list[l] <<" \t\t "<< wilson[l]<<" \t "<<dwf4[l]<<" \t "<<dwf5[l] <<std::endl;
+  }
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+  }
+
+  int NN=NN_global;
+  if ( do_memory ) {
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << " Memory benchmark " <<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    Benchmark::Memory();
+  }
+
+  if ( do_comms && (NN>1) ) {
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << " Communications benchmark " <<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    Benchmark::Comms();
+  }
+
+  if ( do_dwf ) {
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << " Per Node Summary table Ls="<<Ls <<std::endl;
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << " L \t\t Wilson\t\t DWF4  \t\t DWF5 " <<std::endl;
+  for(int l=0;l<L_list.size();l++){
+    std::cout<<GridLogMessage << L_list[l] <<" \t\t "<< wilson[l]/NN<<" \t "<<dwf4[l]/NN<<" \t "<<dwf5[l] /NN<<std::endl;
+  }
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << " Comparison point     result: "  << dwf4[sel]/NN << " Mflop/s per node"<<std::endl;
+  std::cout<<std::setprecision(3);
+  std::cout<<GridLogMessage << " Comparison point robustness: "  << robust_list[sel] <<std::endl;
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+  }
+
+
+  Grid_finalize();
+}

benchmarks/Benchmark_comms.cc

@@ -31,6 +31,32 @@ using namespace std;
 using namespace Grid;
 using namespace Grid::QCD;
 
+struct time_statistics{
+  double mean;
+  double err;
+  double min;
+  double max;
+
+  void statistics(std::vector<double> v){
+      double sum = std::accumulate(v.begin(), v.end(), 0.0);
+      mean = sum / v.size();
+
+      std::vector<double> diff(v.size());
+      std::transform(v.begin(), v.end(), diff.begin(), [=](double x) { return x - mean; });
+      double sq_sum = std::inner_product(diff.begin(), diff.end(), diff.begin(), 0.0);
+      err = std::sqrt(sq_sum / (v.size()*(v.size() - 1)));
+
+      auto result = std::minmax_element(v.begin(), v.end());
+      min = *result.first;
+      max = *result.second;
+}
+};
+
+void header(){
+  std::cout <<GridLogMessage << " L  "<<"\t"<<" Ls  "<<"\t"
+            <<std::setw(11)<<"bytes"<<"MB/s uni (err/min/max)"<<"\t\t"<<"MB/s bidi (err/min/max)"<<std::endl;
+};
+
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
@@ -40,18 +66,21 @@ int main (int argc, char ** argv)
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
-  int Nloop=10;
+  int Nloop=100;
   int nmu=0;
-  for(int mu=0;mu<4;mu++) if (mpi_layout[mu]>1) nmu++;
+  int maxlat=32;
+  for(int mu=0;mu<Nd;mu++) if (mpi_layout[mu]>1) nmu++;
+
+  std::cout << GridLogMessage << "Number of iterations to average: "<< Nloop << std::endl;
+  std::vector<double> t_time(Nloop);
+  time_statistics timestat;
+
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= Benchmarking concurrent halo exchange in "<<nmu<<" dimensions"<<std::endl;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
-  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<" Ls  "<<"\t\t"<<"bytes"<<"\t\t"<<"MB/s uni"<<"\t\t"<<"MB/s bidi"<<std::endl;
-
-
-
-  for(int lat=4;lat<=32;lat+=2){
-    for(int Ls=1;Ls<=16;Ls*=2){
+  header();
+  for(int lat=4;lat<=maxlat;lat+=4){
+    for(int Ls=8;Ls<=8;Ls*=2){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],
       				    lat*mpi_layout[1],
@@ -59,15 +88,23 @@ int main (int argc, char ** argv)
       				    lat*mpi_layout[3]});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      RealD Nrank = Grid._Nprocessors;
+      RealD Nnode = Grid.NodeCount();
+      RealD ppn = Nrank/Nnode;
 
-      std::vector<std::vector<HalfSpinColourVectorD> > xbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
-      std::vector<std::vector<HalfSpinColourVectorD> > rbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
+      std::vector<Vector<HalfSpinColourVectorD> > xbuf(8);	
+      std::vector<Vector<HalfSpinColourVectorD> > rbuf(8);
 
       int ncomm;
       int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
+      for(int mu=0;mu<8;mu++){
+	xbuf[mu].resize(lat*lat*lat*Ls);
+	rbuf[mu].resize(lat*lat*lat*Ls);
+	//	std::cout << " buffers " << std::hex << (uint64_t)&xbuf[mu][0] <<" " << (uint64_t)&rbuf[mu][0] <<std::endl;
+      }
 
-      double start=usecond();
       for(int i=0;i<Nloop;i++){
+      double start=usecond();
 
 	std::vector<CartesianCommunicator::CommsRequest_t> requests;
 
@@ -80,7 +117,6 @@ int main (int argc, char ** argv)
 	    int comm_proc=1;
 	    int xmit_to_rank;
 	    int recv_from_rank;
-	    
 	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
 	    Grid.SendToRecvFromBegin(requests,
 				   (void *)&xbuf[mu][0],
@@ -103,18 +139,24 @@ int main (int argc, char ** argv)
 	}
 	Grid.SendToRecvFromComplete(requests);
 	Grid.Barrier();
-
+	double stop=usecond();
+	t_time[i] = stop-start; // microseconds
       }
-      double stop=usecond();
 
-      double dbytes    = bytes;
-      double xbytes    = Nloop*dbytes*2.0*ncomm;
+      timestat.statistics(t_time);
+
+      double dbytes    = bytes*ppn;
+      double xbytes    = dbytes*2.0*ncomm;
       double rbytes    = xbytes;
       double bidibytes = xbytes+rbytes;
 
-      double time = stop-start; // microseconds
+      std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
+               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
+               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
+               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
+               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
+               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
 
-      std::cout<<GridLogMessage << lat<<"\t\t"<<Ls<<"\t\t"<<bytes<<"\t\t"<<xbytes/time<<"\t\t"<<bidibytes/time<<std::endl;
     }
   }    
 
@@ -122,25 +164,32 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= Benchmarking sequential halo exchange in "<<nmu<<" dimensions"<<std::endl;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
-  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<" Ls  "<<"\t\t"<<"bytes"<<"\t\t"<<"MB/s uni"<<"\t\t"<<"MB/s bidi"<<std::endl;
+  header();
 
-
-  for(int lat=4;lat<=32;lat+=2){
-    for(int Ls=1;Ls<=16;Ls*=2){
+  for(int lat=4;lat<=maxlat;lat+=4){
+    for(int Ls=8;Ls<=8;Ls*=2){
 
       std::vector<int> latt_size  ({lat,lat,lat,lat});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      RealD Nrank = Grid._Nprocessors;
+      RealD Nnode = Grid.NodeCount();
+      RealD ppn = Nrank/Nnode;
 
-      std::vector<std::vector<HalfSpinColourVectorD> > xbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
-      std::vector<std::vector<HalfSpinColourVectorD> > rbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
+      std::vector<Vector<HalfSpinColourVectorD> > xbuf(8);
+      std::vector<Vector<HalfSpinColourVectorD> > rbuf(8);
 
+      for(int mu=0;mu<8;mu++){
+	xbuf[mu].resize(lat*lat*lat*Ls);
+	rbuf[mu].resize(lat*lat*lat*Ls);
+	//	std::cout << " buffers " << std::hex << (uint64_t)&xbuf[mu][0] <<" " << (uint64_t)&rbuf[mu][0] <<std::endl;
+      }
 
       int ncomm;
       int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
 
-      double start=usecond();
       for(int i=0;i<Nloop;i++){
+      double start=usecond();
     
 	ncomm=0;
 	for(int mu=0;mu<4;mu++){
@@ -179,143 +228,314 @@ int main (int argc, char ** argv)
 	  }
 	}
 	Grid.Barrier();
+	double stop=usecond();
+	t_time[i] = stop-start; // microseconds
+
       }
 
-      double stop=usecond();
+      timestat.statistics(t_time);
       
-      double dbytes    = bytes;
-      double xbytes    = Nloop*dbytes*2.0*ncomm;
+      double dbytes    = bytes*ppn;
+      double xbytes    = dbytes*2.0*ncomm;
       double rbytes    = xbytes;
       double bidibytes = xbytes+rbytes;
 
-      double time = stop-start;
+    std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
+               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
+               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
+               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
+               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
+               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
 
-      std::cout<<GridLogMessage << lat<<"\t\t"<<Ls<<"\t\t"<<bytes<<"\t\t"<<xbytes/time<<"\t\t"<<bidibytes/time<<std::endl;
+      
     }
   }  
 
-#if 0
 
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
-  std::cout<<GridLogMessage << "= Benchmarking sequential persistent halo exchange in "<<nmu<<" dimensions"<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking concurrent STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
-  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<" Ls  "<<"\t\t"<<"bytes"<<"\t\t"<<"MB/s uni"<<"\t\t"<<"MB/s bidi"<<std::endl;
+  header();
 
+  for(int lat=4;lat<=maxlat;lat+=4){
+    for(int Ls=8;Ls<=8;Ls*=2){
 
-  for(int lat=4;lat<=32;lat+=2){
-    for(int Ls=1;Ls<=16;Ls*=2){
-
-      std::vector<int> latt_size  ({lat,lat,lat,lat});
+      std::vector<int> latt_size  ({lat*mpi_layout[0],
+      				    lat*mpi_layout[1],
+      				    lat*mpi_layout[2],
+      				    lat*mpi_layout[3]});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      RealD Nrank = Grid._Nprocessors;
+      RealD Nnode = Grid.NodeCount();
+      RealD ppn = Nrank/Nnode;
 
-      std::vector<std::vector<HalfSpinColourVectorD> > xbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
-      std::vector<std::vector<HalfSpinColourVectorD> > rbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
-
+      std::vector<HalfSpinColourVectorD *> xbuf(8);
+      std::vector<HalfSpinColourVectorD *> rbuf(8);
+      Grid.ShmBufferFreeAll();
+      for(int d=0;d<8;d++){
+	xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	bzero((void *)xbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	bzero((void *)rbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+      }
 
       int ncomm;
       int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
 
+      double dbytes;
+      for(int i=0;i<Nloop;i++){
+	double start=usecond();
 
-      std::vector<CartesianCommunicator::CommsRequest_t> empty;
-      std::vector<std::vector<CartesianCommunicator::CommsRequest_t> > requests_fwd(Nd,empty);
-      std::vector<std::vector<CartesianCommunicator::CommsRequest_t> > requests_bwd(Nd,empty);
-
-      for(int mu=0;mu<4;mu++){
+	dbytes=0;
 	ncomm=0;
-	if (mpi_layout[mu]>1 ) {
-	  ncomm++;
 
-	  int comm_proc;
-	  int xmit_to_rank;
-	  int recv_from_rank;
+	std::vector<CartesianCommunicator::CommsRequest_t> requests;
 
-	  comm_proc=1;
-	  Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
-	  Grid.SendToRecvFromInit(requests_fwd[mu],
-				  (void *)&xbuf[mu][0],
-				  xmit_to_rank,
-				  (void *)&rbuf[mu][0],
-				  recv_from_rank,
-				  bytes);
+	for(int mu=0;mu<4;mu++){
+	
 
-	  comm_proc = mpi_layout[mu]-1;
-	  Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
-	  Grid.SendToRecvFromInit(requests_bwd[mu],
-				  (void *)&xbuf[mu+4][0],
-				  xmit_to_rank,
-				  (void *)&rbuf[mu+4][0],
-				  recv_from_rank,
-				  bytes);
-
-	}
-      }
-
-      {
-	double start=usecond();
-	for(int i=0;i<Nloop;i++){
+	  if (mpi_layout[mu]>1 ) {
 	  
-	  for(int mu=0;mu<4;mu++){
-	    
-	    if (mpi_layout[mu]>1 ) {
-	      
-	      Grid.SendToRecvFromBegin(requests_fwd[mu]);
-	      Grid.SendToRecvFromComplete(requests_fwd[mu]);
-	      Grid.SendToRecvFromBegin(requests_bwd[mu]);
-	      Grid.SendToRecvFromComplete(requests_bwd[mu]);
-	    }
-	  }
-	  Grid.Barrier();
-	}
+	    ncomm++;
+	    int comm_proc=1;
+	    int xmit_to_rank;
+	    int recv_from_rank;
+	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	    dbytes+=
+	      Grid.StencilSendToRecvFromBegin(requests,
+					      (void *)&xbuf[mu][0],
+					      xmit_to_rank,
+					      (void *)&rbuf[mu][0],
+					      recv_from_rank,
+					      bytes,mu);
 	
-	double stop=usecond();
-	
-	double dbytes    = bytes;
-	double xbytes    = Nloop*dbytes*2.0*ncomm;
-	double rbytes    = xbytes;
-	double bidibytes = xbytes+rbytes;
-	
-	double time = stop-start;
-	
-	std::cout<<GridLogMessage << lat<<"\t\t"<<Ls<<"\t\t"<<bytes<<"\t\t"<<xbytes/time<<"\t\t"<<bidibytes/time<<std::endl;
-
-      }
-
-
-      {
-	double start=usecond();
-	for(int i=0;i<Nloop;i++){
+	    comm_proc = mpi_layout[mu]-1;
+	  
+	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	    dbytes+=
+	      Grid.StencilSendToRecvFromBegin(requests,
+					      (void *)&xbuf[mu+4][0],
+					      xmit_to_rank,
+					      (void *)&rbuf[mu+4][0],
+					      recv_from_rank,
+					      bytes,mu+4);
 	  
-	  for(int mu=0;mu<4;mu++){
-	    
-	    if (mpi_layout[mu]>1 ) {
-	      
-	      Grid.SendToRecvFromBegin(requests_fwd[mu]);
-	      Grid.SendToRecvFromBegin(requests_bwd[mu]);
-	      Grid.SendToRecvFromComplete(requests_fwd[mu]);
-	      Grid.SendToRecvFromComplete(requests_bwd[mu]);
-	    }
 	  }
-	  Grid.Barrier();
 	}
-	
+	Grid.StencilSendToRecvFromComplete(requests,0);
+	Grid.Barrier();
 	double stop=usecond();
+	t_time[i] = stop-start; // microseconds
 	
-	double dbytes    = bytes;
-	double xbytes    = Nloop*dbytes*2.0*ncomm;
-	double rbytes    = xbytes;
-	double bidibytes = xbytes+rbytes;
-	
-	double time = stop-start;
-	
-	std::cout<<GridLogMessage << lat<<"\t\t"<<Ls<<"\t\t"<<bytes<<"\t\t"<<xbytes/time<<"\t\t"<<bidibytes/time<<std::endl;
-
       }
 
+      timestat.statistics(t_time);
+
+      dbytes=dbytes*ppn;
+      double xbytes    = dbytes*0.5;
+      double rbytes    = dbytes*0.5;
+      double bidibytes = dbytes;
+
+      std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
+               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
+               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
+               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
+               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
+               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
+
+
     }
-  }
+  }    
 
-#endif
+
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking sequential STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  header();
+
+  for(int lat=4;lat<=maxlat;lat+=4){
+    for(int Ls=8;Ls<=8;Ls*=2){
+
+      std::vector<int> latt_size  ({lat*mpi_layout[0],
+      				    lat*mpi_layout[1],
+      				    lat*mpi_layout[2],
+      				    lat*mpi_layout[3]});
+
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      RealD Nrank = Grid._Nprocessors;
+      RealD Nnode = Grid.NodeCount();
+      RealD ppn = Nrank/Nnode;
+
+      std::vector<HalfSpinColourVectorD *> xbuf(8);
+      std::vector<HalfSpinColourVectorD *> rbuf(8);
+      Grid.ShmBufferFreeAll();
+      for(int d=0;d<8;d++){
+	xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	bzero((void *)xbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	bzero((void *)rbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+      }
+
+      int ncomm;
+      int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
+      double dbytes;
+      for(int i=0;i<Nloop;i++){
+	double start=usecond();
+
+	std::vector<CartesianCommunicator::CommsRequest_t> requests;
+	dbytes=0;
+	ncomm=0;
+	for(int mu=0;mu<4;mu++){
+	
+	  if (mpi_layout[mu]>1 ) {
+	  
+	    ncomm++;
+	    int comm_proc=1;
+	    int xmit_to_rank;
+	    int recv_from_rank;
+	    
+	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	    dbytes+=
+	      Grid.StencilSendToRecvFromBegin(requests,
+					      (void *)&xbuf[mu][0],
+					      xmit_to_rank,
+					      (void *)&rbuf[mu][0],
+					      recv_from_rank,
+					      bytes,mu);
+	    Grid.StencilSendToRecvFromComplete(requests,mu);
+	    requests.resize(0);
+
+	    comm_proc = mpi_layout[mu]-1;
+	  
+	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	    dbytes+=
+	      Grid.StencilSendToRecvFromBegin(requests,
+					      (void *)&xbuf[mu+4][0],
+					      xmit_to_rank,
+					      (void *)&rbuf[mu+4][0],
+					      recv_from_rank,
+					      bytes,mu+4);
+	    Grid.StencilSendToRecvFromComplete(requests,mu+4);
+	    requests.resize(0);
+	  
+	  }
+	}
+	Grid.Barrier();
+	double stop=usecond();
+	t_time[i] = stop-start; // microseconds
+	
+      }
+
+      timestat.statistics(t_time);
+
+      dbytes=dbytes*ppn;
+      double xbytes    = dbytes*0.5;
+      double rbytes    = dbytes*0.5;
+      double bidibytes = dbytes;
+
+
+      std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
+               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
+               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
+               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
+               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
+               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
+ 
+    }
+  }    
+
+
+
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking threaded STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  header();
+
+  for(int lat=4;lat<=maxlat;lat+=4){
+    for(int Ls=8;Ls<=8;Ls*=2){
+
+      std::vector<int> latt_size  ({lat*mpi_layout[0],
+      				    lat*mpi_layout[1],
+      				    lat*mpi_layout[2],
+      				    lat*mpi_layout[3]});
+
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      RealD Nrank = Grid._Nprocessors;
+      RealD Nnode = Grid.NodeCount();
+      RealD ppn = Nrank/Nnode;
+
+      std::vector<HalfSpinColourVectorD *> xbuf(8);
+      std::vector<HalfSpinColourVectorD *> rbuf(8);
+      Grid.ShmBufferFreeAll();
+      for(int d=0;d<8;d++){
+	xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	bzero((void *)xbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	bzero((void *)rbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+      }
+
+      int ncomm;
+      int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
+      double dbytes;
+      for(int i=0;i<Nloop;i++){
+	double start=usecond();
+
+	std::vector<CartesianCommunicator::CommsRequest_t> requests;
+	dbytes=0;
+	ncomm=0;
+
+	parallel_for(int dir=0;dir<8;dir++){
+
+	  double tbytes;
+	  int mu =dir % 4;
+
+	  if (mpi_layout[mu]>1 ) {
+	  
+	    ncomm++;
+	    int xmit_to_rank;
+	    int recv_from_rank;
+	    if ( dir == mu ) { 
+	      int comm_proc=1;
+	      Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	    } else { 
+	      int comm_proc = mpi_layout[mu]-1;
+	      Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	    }
+
+	    tbytes= Grid.StencilSendToRecvFrom((void *)&xbuf[dir][0], xmit_to_rank,
+					       (void *)&rbuf[dir][0], recv_from_rank, bytes,dir);
+
+#pragma omp atomic
+	    dbytes+=tbytes;
+	  }
+	}
+	Grid.Barrier();
+	double stop=usecond();
+	t_time[i] = stop-start; // microseconds
+      }
+
+      timestat.statistics(t_time);
+
+      dbytes=dbytes*ppn;
+      double xbytes    = dbytes*0.5;
+      double rbytes    = dbytes*0.5;
+      double bidibytes = dbytes;
+
+
+      std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
+               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
+               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
+               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
+               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
+               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
+ 
+    }
+  }    
+
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= All done; Bye Bye"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
 
   Grid_finalize();
 }

benchmarks/Benchmark_dwf.cc

@@ -1,28 +1,22 @@
-    /*************************************************************************************
-
+ /*************************************************************************************
     Grid physics library, www.github.com/paboyle/Grid 
-
     Source file: ./benchmarks/Benchmark_dwf.cc
-
     Copyright (C) 2015
 
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+    Author: paboyle <paboyle@ph.ed.ac.uk>
 
     This program is free software; you can redistribute it and/or modify
     it under the terms of the GNU General Public License as published by
     the Free Software Foundation; either version 2 of the License, or
     (at your option) any later version.
-
     This program is distributed in the hope that it will be useful,
     but WITHOUT ANY WARRANTY; without even the implied warranty of
     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
     GNU General Public License for more details.
-
     You should have received a copy of the GNU General Public License along
     with this program; if not, write to the Free Software Foundation, Inc.,
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
     See the full license in the file "LICENSE" in the top level distribution directory
     *************************************************************************************/
     /*  END LEGAL */
@@ -37,26 +31,21 @@ struct scal {
   d internal;
 };
 
-  Gamma::GammaMatrix Gmu [] = {
-    Gamma::GammaX,
-    Gamma::GammaY,
-    Gamma::GammaZ,
-    Gamma::GammaT
+  Gamma::Algebra Gmu [] = {
+    Gamma::Algebra::GammaX,
+    Gamma::Algebra::GammaY,
+    Gamma::Algebra::GammaZ,
+    Gamma::Algebra::GammaT
   };
 
-bool overlapComms = false;
 typedef WilsonFermion5D<DomainWallVec5dImplR> WilsonFermion5DR;
 typedef WilsonFermion5D<DomainWallVec5dImplF> WilsonFermion5DF;
 typedef WilsonFermion5D<DomainWallVec5dImplD> WilsonFermion5DD;
 
-
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  if( GridCmdOptionExists(argv,argv+argc,"--asynch") ){
-    overlapComms = true;
-  }
 
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
@@ -76,35 +65,66 @@ int main (int argc, char ** argv)
 
   std::vector<int> seeds4({1,2,3,4});
   std::vector<int> seeds5({5,6,7,8});
-
+  
+  std::cout << GridLogMessage << "Initialising 4d RNG" << std::endl;
   GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
+  std::cout << GridLogMessage << "Initialising 5d RNG" << std::endl;
   GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
+  std::cout << GridLogMessage << "Initialised RNGs" << std::endl;
 
   LatticeFermion src   (FGrid); random(RNG5,src);
+#if 0
+  src = zero;
+  {
+    std::vector<int> origin({0,0,0,latt4[2]-1,0});
+    SpinColourVectorF tmp;
+    tmp=zero;
+    tmp()(0)(0)=Complex(-2.0,0.0);
+    std::cout << " source site 0 " << tmp<<std::endl;
+    pokeSite(tmp,src,origin);
+  }
+#else
+  RealD N2 = 1.0/::sqrt(norm2(src));
+  src = src*N2;
+#endif
+
+
   LatticeFermion result(FGrid); result=zero;
   LatticeFermion    ref(FGrid);    ref=zero;
   LatticeFermion    tmp(FGrid);
   LatticeFermion    err(FGrid);
 
+  std::cout << GridLogMessage << "Drawing gauge field" << std::endl;
   LatticeGaugeField Umu(UGrid); 
-  random(RNG4,Umu);
-
-  LatticeGaugeField Umu5d(FGrid); 
+  SU3::HotConfiguration(RNG4,Umu); 
+  std::cout << GridLogMessage << "Random gauge initialised " << std::endl;
+#if 0
+  Umu=1.0;
+  for(int mu=0;mu<Nd;mu++){
+    LatticeColourMatrix ttmp(UGrid);
+    ttmp = PeekIndex<LorentzIndex>(Umu,mu);
+    //    if (mu !=2 ) ttmp = 0;
+    //    ttmp = ttmp* pow(10.0,mu);
+    PokeIndex<LorentzIndex>(Umu,ttmp,mu);
+  }
+  std::cout << GridLogMessage << "Forced to diagonal " << std::endl;
+#endif
 
+  ////////////////////////////////////
+  // Naive wilson implementation
+  ////////////////////////////////////
   // replicate across fifth dimension
+  LatticeGaugeField Umu5d(FGrid); 
+  std::vector<LatticeColourMatrix> U(4,FGrid);
   for(int ss=0;ss<Umu._grid->oSites();ss++){
     for(int s=0;s<Ls;s++){
       Umu5d._odata[Ls*ss+s] = Umu._odata[ss];
     }
   }
-
-  ////////////////////////////////////
-  // Naive wilson implementation
-  ////////////////////////////////////
-  std::vector<LatticeColourMatrix> U(4,FGrid);
   for(int mu=0;mu<Nd;mu++){
     U[mu] = PeekIndex<LorentzIndex>(Umu5d,mu);
   }
+  std::cout << GridLogMessage << "Setting up Cshift based reference " << std::endl;
 
   if (1)
   {
@@ -125,19 +145,32 @@ int main (int argc, char ** argv)
   RealD M5  =1.8;
 
   RealD NP = UGrid->_Nprocessors;
+  RealD NN = UGrid->NodeCount();
 
-  for(int doasm=1;doasm<2;doasm++){
-
-    QCD::WilsonKernelsStatic::AsmOpt=doasm;
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+  std::cout << GridLogMessage<< "* Kernel options --dslash-generic, --dslash-unroll, --dslash-asm" <<std::endl;
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+  std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionR::Dhop                  "<<std::endl;
+  std::cout << GridLogMessage<< "* Vectorising space-time by "<<vComplex::Nsimd()<<std::endl;
+  if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
+  if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
+#ifdef GRID_OMP
+  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
+  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
+#endif
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
 
   DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-  
-  std::cout<<GridLogMessage << "Naive wilson implementation "<<std::endl;
-  std::cout << GridLogMessage<< "Calling Dw"<<std::endl;
-  int ncall =100;
+  int ncall =500;
   if (1) {
-
+    FGrid->Barrier();
     Dw.ZeroCounters();
+    Dw.Dhop(src,result,0);
+    std::cout<<GridLogMessage<<"Called warmup"<<std::endl;
     double t0=usecond();
     for(int i=0;i<ncall;i++){
       __SSC_START;
@@ -145,95 +178,151 @@ int main (int argc, char ** argv)
       __SSC_STOP;
     }
     double t1=usecond();
+    FGrid->Barrier();
     
     double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
     double flops=1344*volume*ncall;
 
     std::cout<<GridLogMessage << "Called Dw "<<ncall<<" times in "<<t1-t0<<" us"<<std::endl;
-    std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
-    std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<std::endl;
+    //    std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
+    //    std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<std::endl;
     std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;
-    std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NP<<std::endl;
+    std::cout<<GridLogMessage << "mflop/s per rank =  "<< flops/(t1-t0)/NP<<std::endl;
+    std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NN<<std::endl;
     err = ref-result; 
     std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
+
+    /*
+    if(( norm2(err)>1.0e-4) ) { 
+      std::cout << "RESULT\n " << result<<std::endl;
+      std::cout << "REF   \n " << ref   <<std::endl;
+      std::cout << "ERR   \n " << err   <<std::endl;
+      FGrid->Barrier();
+      exit(-1);
+    }
+    */
+    assert (norm2(err)< 1.0e-4 );
     Dw.Report();
   }
 
+  DomainWallFermionRL DwH(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  if (1) {
+    FGrid->Barrier();
+    DwH.ZeroCounters();
+    DwH.Dhop(src,result,0);
+    double t0=usecond();
+    for(int i=0;i<ncall;i++){
+      __SSC_START;
+      DwH.Dhop(src,result,0);
+      __SSC_STOP;
+    }
+    double t1=usecond();
+    FGrid->Barrier();
+    
+    double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
+    double flops=1344*volume*ncall;
+
+    std::cout<<GridLogMessage << "Called half prec comms Dw "<<ncall<<" times in "<<t1-t0<<" us"<<std::endl;
+    std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;
+    std::cout<<GridLogMessage << "mflop/s per rank =  "<< flops/(t1-t0)/NP<<std::endl;
+    std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NN<<std::endl;
+    err = ref-result; 
+    std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
+
+    assert (norm2(err)< 1.0e-3 );
+    DwH.Report();
+  }
+
   if (1)
   {
+
+    std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
+    std::cout << GridLogMessage<< "* Benchmarking WilsonFermion5D<DomainWallVec5dImplR>::Dhop "<<std::endl;
+    std::cout << GridLogMessage<< "* Vectorising fifth dimension by "<<vComplex::Nsimd()<<std::endl;
+    if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
+    if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
+#ifdef GRID_OMP
+  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
+  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
+#endif
+    if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
+    if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
+    if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
+    std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
+
     typedef WilsonFermion5D<DomainWallVec5dImplR> WilsonFermion5DR;
     LatticeFermion ssrc(sFGrid);
     LatticeFermion sref(sFGrid);
     LatticeFermion sresult(sFGrid);
 
     WilsonFermion5DR sDw(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,M5);
-  
-    for(int x=0;x<latt4[0];x++){
-    for(int y=0;y<latt4[1];y++){
-    for(int z=0;z<latt4[2];z++){
-    for(int t=0;t<latt4[3];t++){
-    for(int s=0;s<Ls;s++){
-      std::vector<int> site({s,x,y,z,t});
-      SpinColourVector tmp;
-      peekSite(tmp,src,site);
-      pokeSite(tmp,ssrc,site);
-    }}}}}
+
+    localConvert(src,ssrc);
     std::cout<<GridLogMessage<< "src norms "<< norm2(src)<<" " <<norm2(ssrc)<<std::endl;
-    double t0=usecond();
+    FGrid->Barrier();
+    sDw.Dhop(ssrc,sresult,0);
     sDw.ZeroCounters();
+    double t0=usecond();
     for(int i=0;i<ncall;i++){
       __SSC_START;
       sDw.Dhop(ssrc,sresult,0);
       __SSC_STOP;
     }
     double t1=usecond();
+    FGrid->Barrier();
     double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
     double flops=1344*volume*ncall;
 
     std::cout<<GridLogMessage << "Called Dw s_inner "<<ncall<<" times in "<<t1-t0<<" us"<<std::endl;
     std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;
-    std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NP<<std::endl;
+    std::cout<<GridLogMessage << "mflop/s per rank =  "<< flops/(t1-t0)/NP<<std::endl;
+    std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NN<<std::endl;
+    //    std::cout<<GridLogMessage<< "res norms "<< norm2(result)<<" " <<norm2(sresult)<<std::endl;
     sDw.Report();
-  
-    if(0){
-      for(int i=0;i< PerformanceCounter::NumTypes(); i++ ){
-  sDw.Dhop(ssrc,sresult,0);
-  PerformanceCounter Counter(i);
-  Counter.Start();
-  sDw.Dhop(ssrc,sresult,0);
-  Counter.Stop();
-  Counter.Report();
-      }
+    RealD sum=0;
+
+    err=zero;
+    localConvert(sresult,err);
+    err = err - ref;
+    sum = norm2(err);
+    std::cout<<GridLogMessage<<" difference between normal ref and simd is "<<sum<<std::endl;
+    if(sum > 1.0e-4 ){
+      std::cout<< "sD REF\n " <<ref << std::endl;
+      std::cout<< "sD ERR   \n " <<err  <<std::endl;
     }
+    //    assert(sum < 1.0e-4);
 
-    std::cout<<GridLogMessage<< "res norms "<< norm2(result)<<" " <<norm2(sresult)<<std::endl;
+    err=zero;
+    localConvert(sresult,err);
+    err = err - result;
+    sum = norm2(err);
+    std::cout<<GridLogMessage<<" difference between normal result and simd is "<<sum<<std::endl;
+    if(sum > 1.0e-4 ){
+      std::cout<< "sD REF\n " <<result << std::endl;
+      std::cout<< "sD ERR   \n " << err  <<std::endl;
+    }
+    assert(sum < 1.0e-4);
 
-
-    RealF sum=0;
-    for(int x=0;x<latt4[0];x++){
-    for(int y=0;y<latt4[1];y++){
-    for(int z=0;z<latt4[2];z++){
-    for(int t=0;t<latt4[3];t++){
-    for(int s=0;s<Ls;s++){
-      std::vector<int> site({s,x,y,z,t});
-      SpinColourVector normal, simd;
-      peekSite(normal,result,site);
-      peekSite(simd,sresult,site);
-      sum=sum+norm2(normal-simd);
-      if (norm2(normal-simd) > 1.0e-6 ) {
-  std::cout << "site "<<x<<","<<y<<","<<z<<","<<t<<","<<s<<" "<<norm2(normal-simd)<<std::endl;
-  std::cout << "site "<<x<<","<<y<<","<<z<<","<<t<<","<<s<<" normal "<<normal<<std::endl;
-  std::cout << "site "<<x<<","<<y<<","<<z<<","<<t<<","<<s<<" simd   "<<simd<<std::endl;
-      }
-    }}}}}
-    std::cout<<GridLogMessage<<" difference between normal and simd is "<<sum<<std::endl;
-
-
-    if (1) {
+    
+    if(1){
+      std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
+      std::cout << GridLogMessage<< "* Benchmarking WilsonFermion5D<DomainWallVec5dImplR>::DhopEO "<<std::endl;
+      std::cout << GridLogMessage<< "* Vectorising fifth dimension by "<<vComplex::Nsimd()<<std::endl;
+      if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
+      if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
+#ifdef GRID_OMP
+  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
+  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
+#endif
+      if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) 
+	std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
+      if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) 
+	std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
+      if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) 
+	std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
+      std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
 
       LatticeFermion sr_eo(sFGrid);
-      LatticeFermion serr(sFGrid);
-
       LatticeFermion ssrc_e (sFrbGrid);
       LatticeFermion ssrc_o (sFrbGrid);
       LatticeFermion sr_e   (sFrbGrid);
@@ -241,29 +330,30 @@ int main (int argc, char ** argv)
 
       pickCheckerboard(Even,ssrc_e,ssrc);
       pickCheckerboard(Odd,ssrc_o,ssrc);
-
-      setCheckerboard(sr_eo,ssrc_o);
-      setCheckerboard(sr_eo,ssrc_e);
-      serr = sr_eo-ssrc; 
-      std::cout<<GridLogMessage << "EO src norm diff   "<< norm2(serr)<<std::endl;
+      //      setCheckerboard(sr_eo,ssrc_o);
+      //      setCheckerboard(sr_eo,ssrc_e);
 
       sr_e = zero;
       sr_o = zero;
 
+      FGrid->Barrier();
+      sDw.DhopEO(ssrc_o, sr_e, DaggerNo);
       sDw.ZeroCounters();
-      sDw.stat.init("DhopEO");
+      //      sDw.stat.init("DhopEO");
       double t0=usecond();
       for (int i = 0; i < ncall; i++) {
         sDw.DhopEO(ssrc_o, sr_e, DaggerNo);
       }
       double t1=usecond();
-      sDw.stat.print();
+      FGrid->Barrier();
+      //      sDw.stat.print();
 
       double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
       double flops=(1344.0*volume*ncall)/2;
 
       std::cout<<GridLogMessage << "sDeo mflop/s =   "<< flops/(t1-t0)<<std::endl;
-      std::cout<<GridLogMessage << "sDeo mflop/s per node   "<< flops/(t1-t0)/NP<<std::endl;
+      std::cout<<GridLogMessage << "sDeo mflop/s per rank   "<< flops/(t1-t0)/NP<<std::endl;
+      std::cout<<GridLogMessage << "sDeo mflop/s per node   "<< flops/(t1-t0)/NN<<std::endl;
       sDw.Report();
 
       sDw.DhopEO(ssrc_o,sr_e,DaggerNo);
@@ -272,42 +362,75 @@ int main (int argc, char ** argv)
 
       pickCheckerboard(Even,ssrc_e,sresult);
       pickCheckerboard(Odd ,ssrc_o,sresult);
+
       ssrc_e = ssrc_e - sr_e;
+      RealD error = norm2(ssrc_e);
       std::cout<<GridLogMessage << "sE norm diff   "<< norm2(ssrc_e)<< "  vec nrm"<<norm2(sr_e) <<std::endl;
+
       ssrc_o = ssrc_o - sr_o;
+      error+= norm2(ssrc_o);
       std::cout<<GridLogMessage << "sO norm diff   "<< norm2(ssrc_o)<< "  vec nrm"<<norm2(sr_o) <<std::endl;
+
+      if(( error>1.0e-4) ) { 
+	setCheckerboard(ssrc,ssrc_o);
+	setCheckerboard(ssrc,ssrc_e);
+	std::cout<< "DIFF\n " <<ssrc << std::endl;
+	setCheckerboard(ssrc,sr_o);
+	setCheckerboard(ssrc,sr_e);
+	std::cout<< "CBRESULT\n " <<ssrc << std::endl;
+	std::cout<< "RESULT\n " <<sresult<< std::endl;
+      }
+      assert(error<1.0e-4);
     }
 
+  if(0){
+    std::cout << "Single cache warm call to sDw.Dhop " <<std::endl;
+    for(int i=0;i< PerformanceCounter::NumTypes(); i++ ){
+      sDw.Dhop(ssrc,sresult,0);
+      PerformanceCounter Counter(i);
+      Counter.Start();
+      sDw.Dhop(ssrc,sresult,0);
+      Counter.Stop();
+      Counter.Report();
+    }
+  }
 
   }
 
+
+
   if (1)
   { // Naive wilson dag implementation
     ref = zero;
     for(int mu=0;mu<Nd;mu++){
 
-      //    ref =  src - Gamma(Gamma::GammaX)* src ; // 1+gamma_x
+      //    ref =  src - Gamma(Gamma::Algebra::GammaX)* src ; // 1+gamma_x
       tmp = U[mu]*Cshift(src,mu+1,1);
       for(int i=0;i<ref._odata.size();i++){
-  ref._odata[i]+= tmp._odata[i] + Gamma(Gmu[mu])*tmp._odata[i]; ;
+	ref._odata[i]+= tmp._odata[i] + Gamma(Gmu[mu])*tmp._odata[i]; ;
       }
 
       tmp =adj(U[mu])*src;
       tmp =Cshift(tmp,mu+1,-1);
       for(int i=0;i<ref._odata.size();i++){
-  ref._odata[i]+= tmp._odata[i] - Gamma(Gmu[mu])*tmp._odata[i]; ;
+	ref._odata[i]+= tmp._odata[i] - Gamma(Gmu[mu])*tmp._odata[i]; ;
       }
     }
     ref = -0.5*ref;
   }
+  //  dump=1;
   Dw.Dhop(src,result,1);
-  std::cout << GridLogMessage << "Naive wilson implementation Dag" << std::endl;
+  std::cout << GridLogMessage << "Compare to naive wilson implementation Dag to verify correctness" << std::endl;
   std::cout<<GridLogMessage << "Called DwDag"<<std::endl;
-  std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
-  std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<std::endl;
+  std::cout<<GridLogMessage << "norm dag result "<< norm2(result)<<std::endl;
+  std::cout<<GridLogMessage << "norm dag ref    "<< norm2(ref)<<std::endl;
   err = ref-result; 
-  std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
-
+  std::cout<<GridLogMessage << "norm dag diff   "<< norm2(err)<<std::endl;
+  if((norm2(err)>1.0e-4)){
+	std::cout<< "DAG RESULT\n "  <<ref     << std::endl;
+	std::cout<< "DAG sRESULT\n " <<result  << std::endl;
+	std::cout<< "DAG ERR   \n "  << err    <<std::endl;
+  }
   LatticeFermion src_e (FrbGrid);
   LatticeFermion src_o (FrbGrid);
   LatticeFermion r_e   (FrbGrid);
@@ -315,26 +438,45 @@ int main (int argc, char ** argv)
   LatticeFermion r_eo  (FGrid);
 
 
-  std::cout<<GridLogMessage << "Calling Deo and Doe"<<std::endl;
+  std::cout<<GridLogMessage << "Calling Deo and Doe and //assert Deo+Doe == Dunprec"<<std::endl;
   pickCheckerboard(Even,src_e,src);
   pickCheckerboard(Odd,src_o,src);
 
   std::cout<<GridLogMessage << "src_e"<<norm2(src_e)<<std::endl;
   std::cout<<GridLogMessage << "src_o"<<norm2(src_o)<<std::endl;
 
+
+  // S-direction is INNERMOST and takes no part in the parity.
+  std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
+  std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionR::DhopEO                "<<std::endl;
+  std::cout << GridLogMessage<< "* Vectorising space-time by "<<vComplex::Nsimd()<<std::endl;
+  if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
+  if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
+#ifdef GRID_OMP
+  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
+  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
+#endif
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
+  std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
   {
     Dw.ZeroCounters();
+    FGrid->Barrier();
+    Dw.DhopEO(src_o,r_e,DaggerNo);
     double t0=usecond();
     for(int i=0;i<ncall;i++){
       Dw.DhopEO(src_o,r_e,DaggerNo);
     }
     double t1=usecond();
+    FGrid->Barrier();
     
     double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
     double flops=(1344.0*volume*ncall)/2;
 
     std::cout<<GridLogMessage << "Deo mflop/s =   "<< flops/(t1-t0)<<std::endl;
-    std::cout<<GridLogMessage << "Deo mflop/s per node   "<< flops/(t1-t0)/NP<<std::endl;
+    std::cout<<GridLogMessage << "Deo mflop/s per rank   "<< flops/(t1-t0)/NP<<std::endl;
+    std::cout<<GridLogMessage << "Deo mflop/s per node   "<< flops/(t1-t0)/NN<<std::endl;
     Dw.Report();
   }
   Dw.DhopEO(src_o,r_e,DaggerNo);
@@ -350,14 +492,20 @@ int main (int argc, char ** argv)
 
   err = r_eo-result; 
   std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
+  if((norm2(err)>1.0e-4)){
+	std::cout<< "Deo RESULT\n " <<r_eo << std::endl;
+	std::cout<< "Deo REF\n " <<result  << std::endl;
+	std::cout<< "Deo ERR   \n " << err <<std::endl;
+  }
 
   pickCheckerboard(Even,src_e,err);
   pickCheckerboard(Odd,src_o,err);
   std::cout<<GridLogMessage << "norm diff even  "<< norm2(src_e)<<std::endl;
   std::cout<<GridLogMessage << "norm diff odd   "<< norm2(src_o)<<std::endl;
 
-
-  }
-
+  assert(norm2(src_e)<1.0e-4);
+  assert(norm2(src_o)<1.0e-4);
   Grid_finalize();
+  exit(0);
 }
+

benchmarks/Benchmark_dwf_ntpf.cc

@@ -1,153 +0,0 @@
-    /*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-    Source file: ./benchmarks/Benchmark_dwf.cc
-
-    Copyright (C) 2015
-
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#include <Grid/Grid.h>
-
-using namespace std;
-using namespace Grid;
-using namespace Grid::QCD;
-
-template<class d>
-struct scal {
-  d internal;
-};
-
-  Gamma::GammaMatrix Gmu [] = {
-    Gamma::GammaX,
-    Gamma::GammaY,
-    Gamma::GammaZ,
-    Gamma::GammaT
-  };
-
-bool overlapComms = false;
-
-
-int main (int argc, char ** argv)
-{
-  Grid_init(&argc,&argv);
-
-  if( GridCmdOptionExists(argv,argv+argc,"--asynch") ){
-    overlapComms = true;
-  }
-
-  int threads = GridThread::GetThreads();
-  std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
-
-  std::vector<int> latt4 = GridDefaultLatt();
-  const int Ls=16;
-  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
-  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
-  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
-  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
-
-  std::vector<int> seeds4({1,2,3,4});
-  std::vector<int> seeds5({5,6,7,8});
-
-  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
-  GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
-
-  LatticeFermion src   (FGrid); random(RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid);    ref=zero;
-  LatticeFermion    tmp(FGrid);
-  LatticeFermion    err(FGrid);
-
-  ColourMatrix cm = Complex(1.0,0.0);
-
-  LatticeGaugeField Umu(UGrid); 
-  random(RNG4,Umu);
-
-  LatticeGaugeField Umu5d(FGrid); 
-
-  // replicate across fifth dimension
-  for(int ss=0;ss<Umu._grid->oSites();ss++){
-    for(int s=0;s<Ls;s++){
-      Umu5d._odata[Ls*ss+s] = Umu._odata[ss];
-    }
-  }
-
-  ////////////////////////////////////
-  // Naive wilson implementation
-  ////////////////////////////////////
-  std::vector<LatticeColourMatrix> U(4,FGrid);
-  for(int mu=0;mu<Nd;mu++){
-    U[mu] = PeekIndex<LorentzIndex>(Umu5d,mu);
-  }
-
-  if (1)
-  {
-    ref = zero;
-    for(int mu=0;mu<Nd;mu++){
-
-      tmp = U[mu]*Cshift(src,mu+1,1);
-      ref=ref + tmp - Gamma(Gmu[mu])*tmp;
-
-      tmp =adj(U[mu])*src;
-      tmp =Cshift(tmp,mu+1,-1);
-      ref=ref + tmp + Gamma(Gmu[mu])*tmp;
-    }
-    ref = -0.5*ref;
-  }
-
-  RealD mass=0.1;
-  RealD M5  =1.8;
-
-  typename DomainWallFermionR::ImplParams params; 
-  params.overlapCommsCompute = overlapComms;
-  
-  RealD NP = UGrid->_Nprocessors;
-
-
-  QCD::WilsonKernelsStatic::AsmOpt=1;
-
-  DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,params);
-  
-  std::cout<<GridLogMessage << "Calling Dw"<<std::endl;
-  int ncall =50;
-  if (1) {
-
-    double t0=usecond();
-    for(int i=0;i<ncall;i++){
-      Dw.Dhop(src,result,0);
-    }
-    double t1=usecond();
-    
-    double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-    double flops=1344*volume*ncall;
-
-    std::cout<<GridLogMessage << "Called Dw "<<ncall<<" times in "<<t1-t0<<" us"<<std::endl;
-    std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
-    std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<std::endl;
-    std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;
-    std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NP<<std::endl;
-    err = ref-result; 
-    std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
-    //    Dw.Report();
-  }
-  Grid_finalize();
-}

benchmarks/Benchmark_dwf_sweep.cc

@@ -37,11 +37,11 @@ struct scal {
   d internal;
 };
 
-  Gamma::GammaMatrix Gmu [] = {
-    Gamma::GammaX,
-    Gamma::GammaY,
-    Gamma::GammaZ,
-    Gamma::GammaT
+  Gamma::Algebra Gmu [] = {
+    Gamma::Algebra::GammaX,
+    Gamma::Algebra::GammaY,
+    Gamma::Algebra::GammaZ,
+    Gamma::Algebra::GammaT
   };
 
 void benchDw(std::vector<int> & L, int Ls, int threads, int report =0 );
@@ -51,24 +51,26 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+  std::cout << GridLogMessage<< "* Kernel options --dslash-generic, --dslash-unroll, --dslash-asm" <<std::endl;
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+
   const int Ls=8;
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
-
-  if ( getenv("ASMOPT") )  {
-    QCD::WilsonKernelsStatic::AsmOpt=1;
-  } else { 
-    QCD::WilsonKernelsStatic::AsmOpt=0;
-  }
-
   std::cout<<GridLogMessage << "=========================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= Benchmarking DWF"<<std::endl;
   std::cout<<GridLogMessage << "=========================================================================="<<std::endl;
   std::cout<<GridLogMessage << "Volume \t\t\tProcs \t Dw \t eoDw \t sDw \t eosDw (Mflop/s)  "<<std::endl;
   std::cout<<GridLogMessage << "=========================================================================="<<std::endl;
 
-  int Lmax=32;
-  int dmin=0;
+  int Lmax=16;
+  int dmin=2;
   if ( getenv("LMAX") ) Lmax=atoi(getenv("LMAX"));
   if ( getenv("DMIN") ) dmin=atoi(getenv("DMIN"));
   for (int L=8;L<=Lmax;L*=2){

benchmarks/Benchmark_memory_asynch.cc

@@ -66,7 +66,8 @@ int main (int argc, char ** argv)
 
     Vec tsum; tsum = zero;
 
-    GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+    GridParallelRNG          pRNG(&Grid);      
+    pRNG.SeedFixedIntegers(std::vector<int>({56,17,89,101}));
 
     std::vector<double> stop(threads);
     Vector<Vec> sum(threads);
@@ -77,8 +78,7 @@ int main (int argc, char ** argv)
     }
 
     double start=usecond();
-PARALLEL_FOR_LOOP
-    for(int t=0;t<threads;t++){
+    parallel_for(int t=0;t<threads;t++){
 
       sum[t] = x[t]._odata[0];
       for(int i=0;i<Nloop;i++){

benchmarks/Benchmark_memory_bandwidth.cc

@@ -55,21 +55,21 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s"<<"\t\t"<<"Gflop/s"<<"\t\t seconds"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
-  uint64_t lmax=44;
-#define NLOOP (1*lmax*lmax*lmax*lmax/vol)
-  for(int lat=4;lat<=lmax;lat+=4){
+  uint64_t lmax=96;
+#define NLOOP (10*lmax*lmax*lmax*lmax/vol)
+  for(int lat=8;lat<=lmax;lat+=8){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
       uint64_t Nloop=NLOOP;
 
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeVec z(&Grid); //random(pRNG,z);
-      LatticeVec x(&Grid); //random(pRNG,x);
-      LatticeVec y(&Grid); //random(pRNG,y);
+      LatticeVec z(&Grid);// random(pRNG,z);
+      LatticeVec x(&Grid);// random(pRNG,x);
+      LatticeVec y(&Grid);// random(pRNG,y);
       double a=2.0;
 
 
@@ -83,7 +83,7 @@ int main (int argc, char ** argv)
       double time = (stop-start)/Nloop*1000;
       
       double flops=vol*Nvec*2;// mul,add
-      double bytes=3*vol*Nvec*sizeof(Real);
+      double bytes=3.0*vol*Nvec*sizeof(Real);
       std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t"<<flops/time<<"\t\t"<<(stop-start)/1000./1000.<<std::endl;
 
     }
@@ -94,17 +94,17 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s"<<"\t\t"<<"Gflop/s"<<"\t\t seconds"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
   
-  for(int lat=4;lat<=lmax;lat+=4){
+  for(int lat=8;lat<=lmax;lat+=8){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeVec z(&Grid); //random(pRNG,z);
-      LatticeVec x(&Grid); //random(pRNG,x);
-      LatticeVec y(&Grid); //random(pRNG,y);
+      LatticeVec z(&Grid);// random(pRNG,z);
+      LatticeVec x(&Grid);// random(pRNG,x);
+      LatticeVec y(&Grid);// random(pRNG,y);
       double a=2.0;
 
       uint64_t Nloop=NLOOP;
@@ -119,7 +119,7 @@ int main (int argc, char ** argv)
       double time = (stop-start)/Nloop*1000;
      
       double flops=vol*Nvec*2;// mul,add
-      double bytes=3*vol*Nvec*sizeof(Real);
+      double bytes=3.0*vol*Nvec*sizeof(Real);
       std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t"<<flops/time<<"\t\t"<<(stop-start)/1000./1000.<<std::endl;
 
     }
@@ -129,20 +129,20 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s"<<"\t\t"<<"Gflop/s"<<"\t\t seconds"<<std::endl;
 
-  for(int lat=4;lat<=lmax;lat+=4){
+  for(int lat=8;lat<=lmax;lat+=8){
 
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       uint64_t Nloop=NLOOP;
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeVec z(&Grid); //random(pRNG,z);
-      LatticeVec x(&Grid); //random(pRNG,x);
-      LatticeVec y(&Grid); //random(pRNG,y);
+      LatticeVec z(&Grid);// random(pRNG,z);
+      LatticeVec x(&Grid);// random(pRNG,x);
+      LatticeVec y(&Grid);// random(pRNG,y);
       RealD a=2.0;
 
 
@@ -154,7 +154,7 @@ int main (int argc, char ** argv)
       double stop=usecond();
       double time = (stop-start)/Nloop*1000;
       
-      double bytes=2*vol*Nvec*sizeof(Real);
+      double bytes=2.0*vol*Nvec*sizeof(Real);
       double flops=vol*Nvec*1;// mul
       std::cout<<GridLogMessage <<std::setprecision(3) << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t"<<flops/time<<"\t\t"<<(stop-start)/1000./1000.<<std::endl;
 
@@ -166,17 +166,17 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s"<<"\t\t"<<"Gflop/s"<<"\t\t seconds"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=4;lat<=lmax;lat+=4){
+  for(int lat=8;lat<=lmax;lat+=8){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       uint64_t Nloop=NLOOP;
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
-      LatticeVec z(&Grid); //random(pRNG,z);
-      LatticeVec x(&Grid); //random(pRNG,x);
-      LatticeVec y(&Grid); //random(pRNG,y);
+      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+      LatticeVec z(&Grid);// random(pRNG,z);
+      LatticeVec x(&Grid);// random(pRNG,x);
+      LatticeVec y(&Grid);// random(pRNG,y);
       RealD a=2.0;
       Real nn;      
       double start=usecond();
@@ -187,7 +187,7 @@ int main (int argc, char ** argv)
       double stop=usecond();
       double time = (stop-start)/Nloop*1000;
       
-      double bytes=vol*Nvec*sizeof(Real);
+      double bytes=1.0*vol*Nvec*sizeof(Real);
       double flops=vol*Nvec*2;// mul,add
       std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"  \t\t"<<bytes/time<<"\t\t"<<flops/time<< "\t\t"<<(stop-start)/1000./1000.<< "\t\t " <<std::endl;
 

benchmarks/Benchmark_mooee.cc

@@ -0,0 +1,222 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./benchmarks/Benchmark_dwf.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  int threads = GridThread::GetThreads();
+  std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+
+  std::vector<int> latt4 = GridDefaultLatt();
+  const int Ls=16;
+  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+
+  std::cout << GridLogMessage << "Making Vec5d innermost grids"<<std::endl;
+  GridCartesian         * sUGrid   = SpaceTimeGrid::makeFourDimDWFGrid(GridDefaultLatt(),GridDefaultMpi());
+  GridRedBlackCartesian * sUrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(sUGrid);
+  GridCartesian         * sFGrid   = SpaceTimeGrid::makeFiveDimDWFGrid(Ls,UGrid);
+  GridRedBlackCartesian * sFrbGrid = SpaceTimeGrid::makeFiveDimDWFRedBlackGrid(Ls,UGrid);
+
+  std::vector<int> seeds4({1,2,3,4});
+  std::vector<int> seeds5({5,6,7,8});
+
+  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
+  std::cout << GridLogMessage << "Seeded"<<std::endl;
+
+  LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(RNG4,Umu);
+
+  std::cout << GridLogMessage << "made random gauge fields"<<std::endl;
+
+  RealD mass=0.1;
+  RealD M5  =1.8;
+  RealD NP = UGrid->_Nprocessors;
+
+
+  if (1)
+  {
+    const int ncall=1000;
+
+    std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
+    std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionR::Dhop "<<std::endl;
+    std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
+
+    GridParallelRNG RNG5(FGrid);
+    LatticeFermion src(FGrid); random(RNG5,src);
+    LatticeFermion result(FGrid);
+
+    DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+    double t0,t1;
+
+    LatticeFermion r_eo(FGrid);
+    LatticeFermion src_e (FrbGrid);
+    LatticeFermion src_o (FrbGrid);
+    LatticeFermion r_e   (FrbGrid);
+    LatticeFermion r_o   (FrbGrid);
+    
+    pickCheckerboard(Even,src_e,src);
+    pickCheckerboard(Odd,src_o,src);
+    
+    setCheckerboard(r_eo,src_o);
+    setCheckerboard(r_eo,src_e);
+    
+    r_e = zero;
+    r_o = zero;
+
+
+#define BENCH_DW(A,in,out)			\
+    Dw.CayleyZeroCounters();			\
+    Dw. A (in,out);				\
+    FGrid->Barrier();				\
+    t0=usecond();				\
+    for(int i=0;i<ncall;i++){			\
+      Dw. A (in,out);				\
+    }						\
+    t1=usecond();				\
+    FGrid->Barrier();				\
+    Dw.CayleyReport();					\
+    std::cout<<GridLogMessage << "Called " #A " "<< (t1-t0)/ncall<<" us"<<std::endl;\
+    std::cout<<GridLogMessage << "******************"<<std::endl;
+
+#define BENCH_ZDW(A,in,out)			\
+    zDw.CayleyZeroCounters();			\
+    zDw. A (in,out);				\
+    FGrid->Barrier();				\
+    t0=usecond();				\
+    for(int i=0;i<ncall;i++){			\
+      zDw. A (in,out);				\
+    }						\
+    t1=usecond();				\
+    FGrid->Barrier();				\
+    zDw.CayleyReport();							\
+    std::cout<<GridLogMessage << "Called ZDw " #A " "<< (t1-t0)/ncall<<" us"<<std::endl;\
+    std::cout<<GridLogMessage << "******************"<<std::endl;
+
+#define BENCH_DW_SSC(A,in,out)			\
+    Dw.CayleyZeroCounters();			\
+    Dw. A (in,out);				\
+    FGrid->Barrier();				\
+    t0=usecond();				\
+    for(int i=0;i<ncall;i++){			\
+      __SSC_START ;				\
+      Dw. A (in,out);				\
+      __SSC_STOP ;				\
+    }						\
+    t1=usecond();				\
+    FGrid->Barrier();				\
+    Dw.CayleyReport();					\
+    std::cout<<GridLogMessage << "Called " #A " "<< (t1-t0)/ncall<<" us"<<std::endl;\
+    std::cout<<GridLogMessage << "******************"<<std::endl;
+
+#define BENCH_DW_MEO(A,in,out)			\
+    Dw.CayleyZeroCounters();			\
+    Dw. A (in,out,0);				\
+    FGrid->Barrier();				\
+    t0=usecond();				\
+    for(int i=0;i<ncall;i++){			\
+      Dw. A (in,out,0);				\
+    }						\
+    t1=usecond();				\
+    FGrid->Barrier();				\
+    Dw.CayleyReport();					\
+    std::cout<<GridLogMessage << "Called " #A " "<< (t1-t0)/ncall<<" us"<<std::endl;\
+    std::cout<<GridLogMessage << "******************"<<std::endl;
+
+    BENCH_DW_MEO(Dhop    ,src,result);
+    BENCH_DW_MEO(DhopEO  ,src_o,r_e);
+    BENCH_DW(Meooe   ,src_o,r_e);
+    BENCH_DW(Mooee   ,src_o,r_o);
+    BENCH_DW(MooeeInv,src_o,r_o);
+
+  }
+
+  if (1)
+  {
+    const int ncall=1000;
+
+    std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
+    std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionVec5dR::Dhop "<<std::endl;
+    std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
+
+    GridParallelRNG RNG5(sFGrid);
+    LatticeFermion src(sFGrid); random(RNG5,src);
+    LatticeFermion sref(sFGrid);
+    LatticeFermion result(sFGrid);
+
+
+    std::cout<<GridLogMessage << "Constructing Vec5D Dw "<<std::endl;
+    DomainWallFermionVec5dR Dw(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5);
+
+    RealD b=1.5;// Scale factor b+c=2, b-c=1
+    RealD c=0.5;
+    std::vector<ComplexD> gamma(Ls,std::complex<double>(1.0,0.0));
+    ZMobiusFermionVec5dR zDw(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,gamma,b,c);
+
+    std::cout<<GridLogMessage << "Calling Dhop "<<std::endl;
+    FGrid->Barrier();
+
+    double t0,t1;
+
+    LatticeFermion r_eo(sFGrid);
+    LatticeFermion src_e (sFrbGrid);
+    LatticeFermion src_o (sFrbGrid);
+    LatticeFermion r_e   (sFrbGrid);
+    LatticeFermion r_o   (sFrbGrid);
+    
+    pickCheckerboard(Even,src_e,src);
+    pickCheckerboard(Odd,src_o,src);
+    
+    setCheckerboard(r_eo,src_o);
+    setCheckerboard(r_eo,src_e);
+    
+    r_e = zero;
+    r_o = zero;
+
+    BENCH_DW_MEO(Dhop    ,src,result);
+    BENCH_DW_MEO(DhopEO  ,src_o,r_e);
+    BENCH_DW_SSC(Meooe   ,src_o,r_e);
+    BENCH_DW(Mooee   ,src_o,r_o);
+    BENCH_DW(MooeeInv,src_o,r_o);
+
+    BENCH_ZDW(Mooee   ,src_o,r_o);
+    BENCH_ZDW(MooeeInv,src_o,r_o);
+
+  }
+
+  Grid_finalize();
+}

benchmarks/Benchmark_staggered.cc

@@ -0,0 +1,134 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./benchmarks/Benchmark_staggered.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  std::vector<int> latt_size   = GridDefaultLatt();
+  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  std::vector<int> mpi_layout  = GridDefaultMpi();
+  GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
+  GridRedBlackCartesian     RBGrid(latt_size,simd_layout,mpi_layout);
+
+  int threads = GridThread::GetThreads();
+  std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+  std::cout<<GridLogMessage << "Grid floating point word size is REALF"<< sizeof(RealF)<<std::endl;
+  std::cout<<GridLogMessage << "Grid floating point word size is REALD"<< sizeof(RealD)<<std::endl;
+  std::cout<<GridLogMessage << "Grid floating point word size is REAL"<< sizeof(Real)<<std::endl;
+
+  std::vector<int> seeds({1,2,3,4});
+  GridParallelRNG          pRNG(&Grid);
+  pRNG.SeedFixedIntegers(seeds);
+  //  pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9});
+
+  typedef typename ImprovedStaggeredFermionR::FermionField FermionField; 
+  typename ImprovedStaggeredFermionR::ImplParams params; 
+
+  FermionField src   (&Grid); random(pRNG,src);
+  FermionField result(&Grid); result=zero;
+  FermionField    ref(&Grid);    ref=zero;
+  FermionField    tmp(&Grid);    tmp=zero;
+  FermionField    err(&Grid);    tmp=zero;
+  LatticeGaugeField Umu(&Grid); random(pRNG,Umu);
+  std::vector<LatticeColourMatrix> U(4,&Grid);
+
+  double volume=1;
+  for(int mu=0;mu<Nd;mu++){
+    volume=volume*latt_size[mu];
+  }  
+
+  // Only one non-zero (y)
+#if 0
+  Umu=zero;
+  Complex cone(1.0,0.0);
+  for(int nn=0;nn<Nd;nn++){
+    random(pRNG,U[nn]);
+    if(1) {
+      if (nn!=2) { U[nn]=zero; std::cout<<GridLogMessage << "zeroing gauge field in dir "<<nn<<std::endl; }
+      //      else       { U[nn]= cone;std::cout<<GridLogMessage << "unit gauge field in dir "<<nn<<std::endl; }
+      else       { std::cout<<GridLogMessage << "random gauge field in dir "<<nn<<std::endl; }
+    }
+    PokeIndex<LorentzIndex>(Umu,U[nn],nn);
+  }
+#endif
+
+  for(int mu=0;mu<Nd;mu++){
+    U[mu] = PeekIndex<LorentzIndex>(Umu,mu);
+  }
+  ref = zero;
+  /*  
+  { // Naive wilson implementation
+    ref = zero;
+    for(int mu=0;mu<Nd;mu++){
+      //    ref =  src + Gamma(Gamma::GammaX)* src ; // 1-gamma_x
+      tmp = U[mu]*Cshift(src,mu,1);
+      for(int i=0;i<ref._odata.size();i++){
+	ref._odata[i]+= tmp._odata[i] - Gamma(Gmu[mu])*tmp._odata[i]; ;
+      }
+
+      tmp =adj(U[mu])*src;
+      tmp =Cshift(tmp,mu,-1);
+      for(int i=0;i<ref._odata.size();i++){
+	ref._odata[i]+= tmp._odata[i] + Gamma(Gmu[mu])*tmp._odata[i]; ;
+      }
+    }
+  }
+  ref = -0.5*ref;
+  */
+
+  RealD mass=0.1;
+  RealD c1=9.0/8.0;
+  RealD c2=-1.0/24.0;
+  RealD u0=1.0;
+  ImprovedStaggeredFermionR Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0,params);
+  
+  std::cout<<GridLogMessage << "Calling Ds"<<std::endl;
+  int ncall=1000;
+  double t0=usecond();
+  for(int i=0;i<ncall;i++){
+    Ds.Dhop(src,result,0);
+  }
+  double t1=usecond();
+  double flops=(16*(3*(6+8+8)) + 15*3*2)*volume*ncall; // == 66*16 +  == 1146
+  
+  std::cout<<GridLogMessage << "Called Ds"<<std::endl;
+  std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
+  std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<std::endl;
+  std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;
+  err = ref-result; 
+  std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
+
+  Grid_finalize();
+}

benchmarks/Benchmark_su3.cc

@@ -35,13 +35,14 @@ using namespace Grid::QCD;
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
+#define LMAX (64)
 
-  int Nloop=1000;
+  int64_t Nloop=20;
 
   std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
   std::vector<int> mpi_layout  = GridDefaultMpi();
 
-  int threads = GridThread::GetThreads();
+  int64_t threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
@@ -50,19 +51,19 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=2;lat<=32;lat+=2){
+  for(int lat=2;lat<=LMAX;lat+=2){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeColourMatrix z(&Grid);// random(pRNG,z);
-      LatticeColourMatrix x(&Grid);// random(pRNG,x);
-      LatticeColourMatrix y(&Grid);// random(pRNG,y);
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
 
       double start=usecond();
-      for(int i=0;i<Nloop;i++){
+      for(int64_t i=0;i<Nloop;i++){
 	x=x*y;
       }
       double stop=usecond();
@@ -82,20 +83,20 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=2;lat<=32;lat+=2){
+  for(int lat=2;lat<=LMAX;lat+=2){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeColourMatrix z(&Grid); //random(pRNG,z);
-      LatticeColourMatrix x(&Grid); //random(pRNG,x);
-      LatticeColourMatrix y(&Grid); //random(pRNG,y);
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
 
       double start=usecond();
-      for(int i=0;i<Nloop;i++){
+      for(int64_t i=0;i<Nloop;i++){
 	z=x*y;
       }
       double stop=usecond();
@@ -113,20 +114,20 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=2;lat<=32;lat+=2){
+  for(int lat=2;lat<=LMAX;lat+=2){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeColourMatrix z(&Grid); //random(pRNG,z);
-      LatticeColourMatrix x(&Grid); //random(pRNG,x);
-      LatticeColourMatrix y(&Grid); //random(pRNG,y);
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
 
       double start=usecond();
-      for(int i=0;i<Nloop;i++){
+      for(int64_t i=0;i<Nloop;i++){
 	mult(z,x,y);
       }
       double stop=usecond();
@@ -144,20 +145,20 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=2;lat<=32;lat+=2){
+  for(int lat=2;lat<=LMAX;lat+=2){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeColourMatrix z(&Grid); //random(pRNG,z);
-      LatticeColourMatrix x(&Grid); //random(pRNG,x);
-      LatticeColourMatrix y(&Grid); //random(pRNG,y);
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
 
       double start=usecond();
-      for(int i=0;i<Nloop;i++){
+      for(int64_t i=0;i<Nloop;i++){
 	mac(z,x,y);
       }
       double stop=usecond();

benchmarks/Benchmark_wilson.cc

@@ -37,11 +37,11 @@ struct scal {
   d internal;
 };
 
-  Gamma::GammaMatrix Gmu [] = {
-    Gamma::GammaX,
-    Gamma::GammaY,
-    Gamma::GammaZ,
-    Gamma::GammaT
+  Gamma::Algebra Gmu [] = {
+    Gamma::Algebra::GammaX,
+    Gamma::Algebra::GammaY,
+    Gamma::Algebra::GammaZ,
+    Gamma::Algebra::GammaT
   };
 
 bool overlapComms = false;
@@ -69,7 +69,7 @@ int main (int argc, char ** argv)
   std::vector<int> seeds({1,2,3,4});
   GridParallelRNG          pRNG(&Grid);
   pRNG.SeedFixedIntegers(seeds);
-  //  pRNG.SeedRandomDevice();
+  //  pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9});
 
   LatticeFermion src   (&Grid); random(pRNG,src);
   LatticeFermion result(&Grid); result=zero;
@@ -106,7 +106,7 @@ int main (int argc, char ** argv)
   { // Naive wilson implementation
     ref = zero;
     for(int mu=0;mu<Nd;mu++){
-      //    ref =  src + Gamma(Gamma::GammaX)* src ; // 1-gamma_x
+      //    ref =  src + Gamma(Gamma::Algebra::GammaX)* src ; // 1-gamma_x
       tmp = U[mu]*Cshift(src,mu,1);
       for(int i=0;i<ref._odata.size();i++){
 	ref._odata[i]+= tmp._odata[i] - Gamma(Gmu[mu])*tmp._odata[i]; ;
@@ -159,7 +159,7 @@ int main (int argc, char ** argv)
     ref = zero;
     for(int mu=0;mu<Nd;mu++){
 
-      //    ref =  src - Gamma(Gamma::GammaX)* src ; // 1+gamma_x
+      //    ref =  src - Gamma(Gamma::Algebra::GammaX)* src ; // 1+gamma_x
       tmp = U[mu]*Cshift(src,mu,1);
       for(int i=0;i<ref._odata.size();i++){
 	ref._odata[i]+= tmp._odata[i] + Gamma(Gmu[mu])*tmp._odata[i]; ;

benchmarks/Benchmark_wilson_sweep.cc

@@ -30,11 +30,11 @@ struct scal {
   d internal;
 };
 
-Gamma::GammaMatrix Gmu [] = {
-  Gamma::GammaX,
-  Gamma::GammaY,
-  Gamma::GammaZ,
-  Gamma::GammaT
+Gamma::Algebra Gmu [] = {
+    Gamma::Algebra::GammaX,
+    Gamma::Algebra::GammaY,
+    Gamma::Algebra::GammaZ,
+    Gamma::Algebra::GammaT
 };
 
 bool overlapComms = false;
@@ -58,6 +58,19 @@ int main (int argc, char ** argv)
   std::vector<int> seeds({1,2,3,4});
   RealD mass = 0.1;
 
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+  std::cout << GridLogMessage<< "* Kernel options --dslash-generic, --dslash-unroll, --dslash-asm" <<std::endl;
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+  std::cout << GridLogMessage<< "* Benchmarking WilsonFermionR::Dhop                  "<<std::endl;
+  std::cout << GridLogMessage<< "* Vectorising space-time by "<<vComplex::Nsimd()<<std::endl;
+  if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
+  if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+
   std::cout<<GridLogMessage << "============================================================================="<< std::endl;
   std::cout<<GridLogMessage << "= Benchmarking Wilson" << std::endl;
   std::cout<<GridLogMessage << "============================================================================="<< std::endl;

benchmarks/Benchmark_zmm.cc

@@ -1,175 +0,0 @@
-    /*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-    Source file: ./tests/Test_zmm.cc
-
-    Copyright (C) 2015
-
-Author: paboyle <paboyle@ph.ed.ac.uk>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#include <Grid/Grid.h>
-
-
-using namespace Grid;
-using namespace Grid::QCD;
-
-
-int bench(std::ofstream &os, std::vector<int> &latt4,int Ls);
-
-int main(int argc,char **argv)
-{
-  Grid_init(&argc,&argv);
-  std::ofstream os("zmm.dat");
-
-  os << "#V Ls Lxy Lzt C++ Asm OMP L1 " <<std::endl;
-  std::cout<<GridLogMessage << "====================================================================="<<std::endl;
-  std::cout<<GridLogMessage << "= Benchmarking ZMM"<<std::endl;
-  std::cout<<GridLogMessage << "====================================================================="<<std::endl;
-  std::cout<<GridLogMessage << "Volume \t\t\t\tC++DW/MFLOPs\tASM-DW/MFLOPs\tdiff"<<std::endl;
-  std::cout<<GridLogMessage << "====================================================================="<<std::endl;
-  for(int L=4;L<=32;L+=4){
-    for(int m=1;m<=2;m++){
-      for(int Ls=8;Ls<=16;Ls+=8){
-	std::vector<int> grid({L,L,m*L,m*L});
-  std::cout << GridLogMessage <<"\t";
-	for(int i=0;i<4;i++) { 
-	  std::cout << grid[i]<<"x";
-	}
-	std::cout << Ls<<"\t\t";
-	bench(os,grid,Ls);
-      }
-    }
-  }
-}
-
-int bench(std::ofstream &os, std::vector<int> &latt4,int Ls)
-{
-
-  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(latt4, GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
-  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
-  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
-  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
-
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
-  int threads = GridThread::GetThreads();
-
-  std::vector<int> seeds4({1,2,3,4});
-  std::vector<int> seeds5({5,6,7,8});
-
-  GridSerialRNG sRNG; sRNG.SeedFixedIntegers(seeds4);
-
-  LatticeFermion src (FGrid);
-  LatticeFermion tmp (FGrid);
-  LatticeFermion srce(FrbGrid);
-
-  LatticeFermion resulto(FrbGrid); resulto=zero;
-  LatticeFermion resulta(FrbGrid); resulta=zero;
-  LatticeFermion junk(FrbGrid); junk=zero;
-  LatticeFermion diff(FrbGrid); 
-  LatticeGaugeField Umu(UGrid);
-
-  double mfc, mfa, mfo, mfl1;
-
-  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
-  GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
-  random(RNG5,src);
-#if 1
-  random(RNG4,Umu);
-#else
-  int mmu=2;
-  std::vector<LatticeColourMatrix> U(4,UGrid);
-  for(int mu=0;mu<Nd;mu++){
-    U[mu] = PeekIndex<LorentzIndex>(Umu,mu);
-    if ( mu!=mmu ) U[mu] = zero;
-    if ( mu==mmu ) U[mu] = 1.0;
-    PokeIndex<LorentzIndex>(Umu,U[mu],mu);
-  }
-#endif
- pickCheckerboard(Even,srce,src);
-
-  RealD mass=0.1;
-  RealD M5  =1.8;
-  DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-
-  int ncall=50;
-  double t0=usecond();
-  for(int i=0;i<ncall;i++){
-    Dw.DhopOE(srce,resulto,0);
-  }
-  double t1=usecond();
-
-  double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-  double flops=1344*volume/2;
-
-  mfc = flops*ncall/(t1-t0);
-  std::cout<<mfc<<"\t\t";
-
-  QCD::WilsonKernelsStatic::AsmOpt=1;
-  t0=usecond();
-  for(int i=0;i<ncall;i++){
-    Dw.DhopOE(srce,resulta,0);
-  }
-  t1=usecond();
-  mfa = flops*ncall/(t1-t0);
-  std::cout<<mfa<<"\t\t";
-  /*
-  int dag=DaggerNo;
-  t0=usecond();
-  for(int i=0;i<1;i++){
-    Dw.DhopInternalOMPbench(Dw.StencilEven,Dw.LebesgueEvenOdd,Dw.UmuOdd,srce,resulta,dag);
-  }
-  t1=usecond();
-  mfo = flops*100/(t1-t0);
-  std::cout<<GridLogMessage << "Called ASM-OMP Dw"<< " mflop/s =   "<< mfo<<std::endl;
-
-  t0=usecond();
-  for(int i=0;i<1;i++){
-    Dw.DhopInternalL1bench(Dw.StencilEven,Dw.LebesgueEvenOdd,Dw.UmuOdd,srce,resulta,dag);
-  }
-  t1=usecond();
-  mfl1= flops*100/(t1-t0);
-  std::cout<<GridLogMessage << "Called ASM-L1 Dw"<< " mflop/s =   "<< mfl1<<std::endl;
-  os << latt4[0]*latt4[1]*latt4[2]*latt4[3]<< " "<<Ls<<" "<< latt4[0] <<" " <<latt4[2]<< " "
-     << mfc<<" "
-     << mfa<<" "
-     << mfo<<" "
-     << mfl1<<std::endl;
-  */
-
-#if 0
-  for(int i=0;i< PerformanceCounter::NumTypes(); i++ ){
-    Dw.DhopOE(srce,resulta,0);
-    PerformanceCounter Counter(i);
-    Counter.Start();
-    Dw.DhopOE(srce,resulta,0);
-    Counter.Stop();
-    Counter.Report();
-  }
-#endif
-  //resulta = (-0.5) * resulta;
-
-  diff = resulto-resulta;
-  std::cout<<norm2(diff)<<std::endl;
-  return 0;
-}
-
-

benchmarks/Makefile.am

@@ -1 +1,7 @@
 include Make.inc
+
+bench-local: all
+	./Benchmark_su3
+	./Benchmark_memory_bandwidth
+	./Benchmark_wilson
+	./Benchmark_dwf --dslash-unroll

benchmarks/simple_simd_test.cc

@@ -0,0 +1,11 @@
+#include <Grid/Grid.h>
+
+Grid::vRealD add(const Grid::vRealD &x, const Grid::vRealD &y)
+{
+	return x+y;
+}
+
+Grid::vRealD sub(const Grid::vRealD &x, const Grid::vRealD &y)
+{
+	return x-y;
+}

benchmarks/simple_su3_expr.cc

@@ -25,7 +25,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     See the full license in the file "LICENSE" in the top level distribution directory
     *************************************************************************************/
     /*  END LEGAL */
-#include <Grid.h>
+#include <Grid/Grid.h>
 
 using namespace std;
 using namespace Grid;

benchmarks/simple_su3_test.cc

@@ -25,7 +25,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     See the full license in the file "LICENSE" in the top level distribution directory
     *************************************************************************************/
     /*  END LEGAL */
-#include <Grid.h>
+#include <Grid/Grid.h>
 
 using namespace std;
 using namespace Grid;

bootstrap.sh

@@ -1,18 +1,12 @@
 #!/usr/bin/env bash
 
-EIGEN_URL='http://bitbucket.org/eigen/eigen/get/3.2.9.tar.bz2'
-FFTW_URL=http://www.fftw.org/fftw-3.3.4.tar.gz
+EIGEN_URL='http://bitbucket.org/eigen/eigen/get/3.3.3.tar.bz2'
 
 echo "-- deploying Eigen source..."
 wget ${EIGEN_URL} --no-check-certificate
 ./scripts/update_eigen.sh `basename ${EIGEN_URL}`
 rm `basename ${EIGEN_URL}`
 
-echo "-- copying fftw prototypes..."
-wget ${FFTW_URL}
-./scripts/update_fftw.sh `basename ${FFTW_URL}`
-rm `basename ${FFTW_URL}`
-
 echo '-- generating Make.inc files...'
 ./scripts/filelist
 echo '-- generating configure script...'

configure.ac

@@ -1,30 +1,50 @@
 AC_PREREQ([2.63])
-AC_INIT([Grid], [0.5.1-dev], [https://github.com/paboyle/Grid], [Grid])
+AC_INIT([Grid], [0.7.0], [https://github.com/paboyle/Grid], [Grid])
 AC_CANONICAL_BUILD
 AC_CANONICAL_HOST
 AC_CANONICAL_TARGET
-AM_INIT_AUTOMAKE(subdir-objects)
+AM_INIT_AUTOMAKE([subdir-objects 1.13])
+AM_EXTRA_RECURSIVE_TARGETS([tests bench])
 AC_CONFIG_MACRO_DIR([m4])
 AC_CONFIG_SRCDIR([lib/Grid.h])
-AC_CONFIG_HEADERS([lib/Config.h])
+AC_CONFIG_HEADERS([lib/Config.h],[sed -i 's|PACKAGE_|GRID_|' lib/Config.h])
 m4_ifdef([AM_SILENT_RULES], [AM_SILENT_RULES([yes])])
 
+################ Get git info
+#AC_REVISION([m4_esyscmd_s([./scripts/configure.commit])])
+
+################ Set flags
+# do not move!
+CXXFLAGS="-O3 $CXXFLAGS"
 
 ############### Checks for programs
-AC_LANG(C++)
-CXXFLAGS="-O3 $CXXFLAGS"
 AC_PROG_CXX
 AC_PROG_RANLIB
 
-############ openmp  ###############
+############### Get compiler informations
+AC_LANG([C++])
+AX_CXX_COMPILE_STDCXX_11([noext],[mandatory])
+AX_COMPILER_VENDOR
+AC_DEFINE_UNQUOTED([CXX_COMP_VENDOR],["$ax_cv_cxx_compiler_vendor"],
+      [vendor of C++ compiler that will compile the code])
+AX_GXX_VERSION
+AC_DEFINE_UNQUOTED([GXX_VERSION],["$GXX_VERSION"],
+      [version of g++ that will compile the code])
+
+
+
+############### Checks for typedefs, structures, and compiler characteristics
+AC_TYPE_SIZE_T
+AC_TYPE_UINT32_T
+AC_TYPE_UINT64_T
+
+############### OpenMP
 AC_OPENMP
-
 ac_openmp=no
-
 if test "${OPENMP_CXXFLAGS}X" != "X"; then
-ac_openmp=yes
-AM_CXXFLAGS="$OPENMP_CXXFLAGS $AM_CXXFLAGS"
-AM_LDFLAGS="$OPENMP_CXXFLAGS $AM_LDFLAGS"
+  ac_openmp=yes
+  AM_CXXFLAGS="$OPENMP_CXXFLAGS $AM_CXXFLAGS"
+  AM_LDFLAGS="$OPENMP_CXXFLAGS $AM_LDFLAGS"
 fi
 
 ############### Checks for header files
@@ -34,15 +54,15 @@ AC_CHECK_HEADERS(malloc/malloc.h)
 AC_CHECK_HEADERS(malloc.h)
 AC_CHECK_HEADERS(endian.h)
 AC_CHECK_HEADERS(execinfo.h)
+AC_CHECK_HEADERS(numaif.h)
 AC_CHECK_DECLS([ntohll],[], [], [[#include <arpa/inet.h>]])
 AC_CHECK_DECLS([be64toh],[], [], [[#include <arpa/inet.h>]])
 
-############### Checks for typedefs, structures, and compiler characteristics
-AC_TYPE_SIZE_T
-AC_TYPE_UINT32_T
-AC_TYPE_UINT64_T
+############## Standard libraries
+AC_CHECK_LIB([m],[cos])
+AC_CHECK_LIB([stdc++],[abort])
 
-############### GMP and MPFR #################
+############### GMP and MPFR
 AC_ARG_WITH([gmp],
     [AS_HELP_STRING([--with-gmp=prefix],
     [try this for a non-standard install prefix of the GMP library])],
@@ -54,10 +74,24 @@ AC_ARG_WITH([mpfr],
     [AM_CXXFLAGS="-I$with_mpfr/include $AM_CXXFLAGS"]
     [AM_LDFLAGS="-L$with_mpfr/lib $AM_LDFLAGS"])
 
-################## lapack ####################
+############### FFTW3
+AC_ARG_WITH([fftw],
+            [AS_HELP_STRING([--with-fftw=prefix],
+            [try this for a non-standard install prefix of the FFTW3 library])],
+            [AM_CXXFLAGS="-I$with_fftw/include $AM_CXXFLAGS"]
+            [AM_LDFLAGS="-L$with_fftw/lib $AM_LDFLAGS"])
+
+############### LIME
+AC_ARG_WITH([lime],
+            [AS_HELP_STRING([--with-lime=prefix],
+            [try this for a non-standard install prefix of the LIME library])],
+            [AM_CXXFLAGS="-I$with_lime/include $AM_CXXFLAGS"]
+            [AM_LDFLAGS="-L$with_lime/lib $AM_LDFLAGS"])
+
+############### lapack
 AC_ARG_ENABLE([lapack],
-    [AC_HELP_STRING([--enable-lapack=yes|no|prefix], [enable LAPACK])], 
-    [ac_LAPACK=${enable_lapack}],[ac_LAPACK=no])
+    [AC_HELP_STRING([--enable-lapack=yes|no|prefix], [enable LAPACK])],
+    [ac_LAPACK=${enable_lapack}], [ac_LAPACK=no])
 
 case ${ac_LAPACK} in
     no)
@@ -67,12 +101,47 @@ case ${ac_LAPACK} in
     *)
         AM_CXXFLAGS="-I$ac_LAPACK/include $AM_CXXFLAGS"
         AM_LDFLAGS="-L$ac_LAPACK/lib $AM_LDFLAGS"
-        AC_DEFINE([USE_LAPACK],[1],[use LAPACK])
+        AC_DEFINE([USE_LAPACK],[1],[use LAPACK]);;
 esac
 
-################## first-touch ####################
+############### FP16 conversions
+AC_ARG_ENABLE([sfw-fp16],
+    [AC_HELP_STRING([--enable-sfw-fp16=yes|no], [enable software fp16 comms])],
+    [ac_SFW_FP16=${enable_sfw_fp16}], [ac_SFW_FP16=yes])
+case ${ac_SFW_FP16} in
+    yes)
+      AC_DEFINE([SFW_FP16],[1],[software conversion to fp16]);;
+    no);;
+    *)
+      AC_MSG_ERROR(["SFW FP16 option not supported ${ac_SFW_FP16}"]);;
+esac
+
+############### MKL
+AC_ARG_ENABLE([mkl],
+    [AC_HELP_STRING([--enable-mkl=yes|no|prefix], [enable Intel MKL for LAPACK & FFTW])],
+    [ac_MKL=${enable_mkl}], [ac_MKL=no])
+
+case ${ac_MKL} in
+    no)
+        ;;
+    yes)
+        AC_DEFINE([USE_MKL], [1], [Define to 1 if you use the Intel MKL]);;
+    *)
+        AM_CXXFLAGS="-I$ac_MKL/include $AM_CXXFLAGS"
+        AM_LDFLAGS="-L$ac_MKL/lib $AM_LDFLAGS"
+        AC_DEFINE([USE_MKL], [1], [Define to 1 if you use the Intel MKL]);;
+esac
+
+############### HDF5
+AC_ARG_WITH([hdf5],
+    [AS_HELP_STRING([--with-hdf5=prefix],
+    [try this for a non-standard install prefix of the HDF5 library])],
+    [AM_CXXFLAGS="-I$with_hdf5/include $AM_CXXFLAGS"]
+    [AM_LDFLAGS="-L$with_hdf5/lib $AM_LDFLAGS"])
+
+############### first-touch
 AC_ARG_ENABLE([numa],
-    [AC_HELP_STRING([--enable-numa=yes|no|prefix], [enable first touch numa opt])], 
+    [AC_HELP_STRING([--enable-numa=yes|no|prefix], [enable first touch numa opt])],
     [ac_NUMA=${enable_NUMA}],[ac_NUMA=no])
 
 case ${ac_NUMA} in
@@ -84,84 +153,118 @@ case ${ac_NUMA} in
         AC_DEFINE([GRID_NUMA],[1],[First touch numa locality]);;
 esac
 
-################## FFTW3 ####################
-AC_ARG_WITH([fftw],    
-            [AS_HELP_STRING([--with-fftw=prefix],
-            [try this for a non-standard install prefix of the FFTW3 library])],
-            [AM_CXXFLAGS="-I$with_fftw/include $AM_CXXFLAGS"]
-            [AM_LDFLAGS="-L$with_fftw/lib $AM_LDFLAGS"])
-
-################ Get compiler informations
-AC_LANG([C++])
-AX_CXX_COMPILE_STDCXX_11([noext],[mandatory])
-AX_COMPILER_VENDOR
-AC_DEFINE_UNQUOTED([CXX_COMP_VENDOR],["$ax_cv_cxx_compiler_vendor"],
-      [vendor of C++ compiler that will compile the code])
-AX_GXX_VERSION
-AC_DEFINE_UNQUOTED([GXX_VERSION],["$GXX_VERSION"],
-      [version of g++ that will compile the code])
-
 ############### Checks for library functions
 CXXFLAGS_CPY=$CXXFLAGS
 LDFLAGS_CPY=$LDFLAGS
 CXXFLAGS="$AM_CXXFLAGS $CXXFLAGS"
 LDFLAGS="$AM_LDFLAGS $LDFLAGS"
+
 AC_CHECK_FUNCS([gettimeofday])
-AC_CHECK_LIB([gmp],[__gmpf_init],
-             [AC_CHECK_LIB([mpfr],[mpfr_init],
-                 [AC_DEFINE([HAVE_LIBMPFR], [1], [Define to 1 if you have the `MPFR' library (-lmpfr).])]
-                 [have_mpfr=true]
-                 [LIBS="$LIBS -lmpfr"],
-                 [AC_MSG_ERROR([MPFR library not found])])]
-   	     [AC_DEFINE([HAVE_LIBGMP], [1], [Define to 1 if you have the `GMP' library (-lgmp).])]
-             [have_gmp=true]
-             [LIBS="$LIBS -lgmp"],
-             [AC_MSG_WARN([**** GMP library not found, Grid can still compile but RHMC will not work ****])])
+
+if test "${ac_MKL}x" != "nox"; then
+    AC_SEARCH_LIBS([mkl_set_interface_layer], [mkl_rt], [],
+                   [AC_MSG_ERROR("MKL enabled but library not found")])
+fi
+
+AC_SEARCH_LIBS([__gmpf_init], [gmp],
+               [AC_SEARCH_LIBS([mpfr_init], [mpfr],
+                               [AC_DEFINE([HAVE_LIBMPFR], [1],
+                                          [Define to 1 if you have the `MPFR' library])]
+                               [have_mpfr=true], [AC_MSG_ERROR([MPFR library not found])])]
+               [AC_DEFINE([HAVE_LIBGMP], [1], [Define to 1 if you have the `GMP' library])]
+               [have_gmp=true])
 
 if test "${ac_LAPACK}x" != "nox"; then
-    AC_CHECK_LIB([lapack],[LAPACKE_sbdsdc],[],
-                 [AC_MSG_ERROR("LAPACK enabled but library not found")])
+    AC_SEARCH_LIBS([LAPACKE_sbdsdc], [lapack], [],
+                   [AC_MSG_ERROR("LAPACK enabled but library not found")])
 fi
-AC_CHECK_LIB([fftw3],[fftw_execute],
-  [AC_DEFINE([HAVE_FFTW],[1],[Define to 1 if you have the `FFTW' library (-lfftw3).])]
-  [have_fftw=true]
-  [LIBS="$LIBS -lfftw3 -lfftw3f"],
-  [AC_MSG_WARN([**** FFTW library not found, Grid can still compile but FFT-based routines will not work ****])])
+
+AC_SEARCH_LIBS([fftw_execute], [fftw3],
+               [AC_SEARCH_LIBS([fftwf_execute], [fftw3f], [],
+                               [AC_MSG_ERROR("single precision FFTW library not found")])]
+               [AC_DEFINE([HAVE_FFTW], [1], [Define to 1 if you have the `FFTW' library])]
+               [have_fftw=true])
+
+AC_SEARCH_LIBS([limeCreateReader], [lime],
+               [AC_DEFINE([HAVE_LIME], [1], [Define to 1 if you have the `LIME' library])]
+               [have_lime=true],
+	       [AC_MSG_WARN(C-LIME library was not found in your system.
+In order to use ILGG file format please install or provide the correct path to your installation
+Info at: http://usqcd.jlab.org/usqcd-docs/c-lime/)])
+
+AC_SEARCH_LIBS([crc32], [z],
+               [AC_DEFINE([HAVE_ZLIB], [1], [Define to 1 if you have the `LIBZ' library])]
+               [have_zlib=true] [LIBS="${LIBS} -lz"],
+	       [AC_MSG_ERROR(zlib library was not found in your system.)])
+
+AC_SEARCH_LIBS([move_pages], [numa],
+               [AC_DEFINE([HAVE_LIBNUMA], [1], [Define to 1 if you have the `LIBNUMA' library])]
+               [have_libnuma=true] [LIBS="${LIBS} -lnuma"],
+	       [AC_MSG_WARN(libnuma library was not found in your system. Some optimisations will not apply)])
+
+AC_SEARCH_LIBS([H5Fopen], [hdf5_cpp],
+               [AC_DEFINE([HAVE_HDF5], [1], [Define to 1 if you have the `HDF5' library])]
+               [have_hdf5=true]
+               [LIBS="${LIBS} -lhdf5"], [], [-lhdf5])
+AM_CONDITIONAL(BUILD_HDF5, [ test "${have_hdf5}X" == "trueX" ])
+
 CXXFLAGS=$CXXFLAGS_CPY
 LDFLAGS=$LDFLAGS_CPY
 
 ############### SIMD instruction selection
-AC_ARG_ENABLE([simd],[AC_HELP_STRING([--enable-simd=SSE4|AVX|AVXFMA4|AVXFMA|AVX2|AVX512|AVX512MIC|IMCI|KNL|KNC],\
-	[Select instructions to be SSE4.0, AVX 1.0, AVX 2.0+FMA, AVX 512, IMCI])],\
-	[ac_SIMD=${enable_simd}],[ac_SIMD=GEN])
+AC_ARG_ENABLE([simd],[AC_HELP_STRING([--enable-simd=code],
+              [select SIMD target (cf. README.md)])], [ac_SIMD=${enable_simd}], [ac_SIMD=GEN])
+
+AC_ARG_ENABLE([gen-simd-width],
+            [AS_HELP_STRING([--enable-gen-simd-width=size],
+            [size (in bytes) of the generic SIMD vectors (default: 32)])],
+            [ac_gen_simd_width=$enable_gen_simd_width],
+            [ac_gen_simd_width=32])
 
 case ${ax_cv_cxx_compiler_vendor} in
   clang|gnu)
     case ${ac_SIMD} in
       SSE4)
         AC_DEFINE([SSE4],[1],[SSE4 intrinsics])
-        SIMD_FLAGS='-msse4.2';;
+	case ${ac_SFW_FP16} in
+	  yes)
+	  SIMD_FLAGS='-msse4.2';;
+	  no)
+	  SIMD_FLAGS='-msse4.2 -mf16c';;
+	  *)
+          AC_MSG_ERROR(["SFW_FP16 must be either yes or no value ${ac_SFW_FP16} "]);;
+	esac;;
       AVX)
         AC_DEFINE([AVX1],[1],[AVX intrinsics])
-        SIMD_FLAGS='-mavx';;
+        SIMD_FLAGS='-mavx -mf16c';;
       AVXFMA4)
         AC_DEFINE([AVXFMA4],[1],[AVX intrinsics with FMA4])
-        SIMD_FLAGS='-mavx -mfma4';;
+        SIMD_FLAGS='-mavx -mfma4 -mf16c';;
       AVXFMA)
         AC_DEFINE([AVXFMA],[1],[AVX intrinsics with FMA3])
-        SIMD_FLAGS='-mavx -mfma';;
+        SIMD_FLAGS='-mavx -mfma -mf16c';;
       AVX2)
         AC_DEFINE([AVX2],[1],[AVX2 intrinsics])
-        SIMD_FLAGS='-mavx2 -mfma';;
-      AVX512|AVX512MIC|KNL)
+        SIMD_FLAGS='-mavx2 -mfma -mf16c';;
+      AVX512)
         AC_DEFINE([AVX512],[1],[AVX512 intrinsics])
         SIMD_FLAGS='-mavx512f -mavx512pf -mavx512er -mavx512cd';;
-      IMCI|KNC)
+      KNC)
         AC_DEFINE([IMCI],[1],[IMCI intrinsics for Knights Corner])
         SIMD_FLAGS='';;
+      KNL)
+        AC_DEFINE([AVX512],[1],[AVX512 intrinsics])
+        AC_DEFINE([KNL],[1],[Knights landing processor])
+        SIMD_FLAGS='-march=knl';;
       GEN)
-        AC_DEFINE([GENERIC_VEC],[1],[generic vector code])
+        AC_DEFINE([GEN],[1],[generic vector code])
+        AC_DEFINE_UNQUOTED([GEN_SIMD_WIDTH],[$ac_gen_simd_width],
+                           [generic SIMD vector width (in bytes)])
+        SIMD_GEN_WIDTH_MSG=" (width= $ac_gen_simd_width)"
         SIMD_FLAGS='';;
+      NEONv8)
+        AC_DEFINE([NEONV8],[1],[ARMv8 NEON])
+        SIMD_FLAGS='-march=armv8-a';;
       QPX|BGQ)
         AC_DEFINE([QPX],[1],[QPX intrinsics for BG/Q])
         SIMD_FLAGS='';;
@@ -176,26 +279,27 @@ case ${ax_cv_cxx_compiler_vendor} in
       AVX)
         AC_DEFINE([AVX1],[1],[AVX intrinsics])
         SIMD_FLAGS='-mavx -xavx';;
-      AVXFMA4)
-        AC_DEFINE([AVXFMA4],[1],[AVX intrinsics with FMA4])
-        SIMD_FLAGS='-mavx -mfma';;
       AVXFMA)
-        AC_DEFINE([AVXFMA],[1],[AVX intrinsics with FMA4])
-        SIMD_FLAGS='-mavx -mfma';;
+        AC_DEFINE([AVXFMA],[1],[AVX intrinsics with FMA3])
+        SIMD_FLAGS='-mavx -fma';;
       AVX2)
         AC_DEFINE([AVX2],[1],[AVX2 intrinsics])
         SIMD_FLAGS='-march=core-avx2 -xcore-avx2';;
       AVX512)
         AC_DEFINE([AVX512],[1],[AVX512 intrinsics])
         SIMD_FLAGS='-xcore-avx512';;
-      AVX512MIC|KNL)
-        AC_DEFINE([AVX512],[1],[AVX512 intrinsics for Knights Landing])
-        SIMD_FLAGS='-xmic-avx512';;
-      IMCI|KNC)
+      KNC)
         AC_DEFINE([IMCI],[1],[IMCI Intrinsics for Knights Corner])
         SIMD_FLAGS='';;
+      KNL)
+        AC_DEFINE([AVX512],[1],[AVX512 intrinsics for Knights Landing])
+        AC_DEFINE([KNL],[1],[Knights landing processor])
+        SIMD_FLAGS='-xmic-avx512';;
       GEN)
-        AC_DEFINE([GENERIC_VEC],[1],[generic vector code])
+        AC_DEFINE([GEN],[1],[generic vector code])
+        AC_DEFINE_UNQUOTED([GEN_SIMD_WIDTH],[$ac_gen_simd_width],
+                           [generic SIMD vector width (in bytes)])
+        SIMD_GEN_WIDTH_MSG=" (width= $ac_gen_simd_width)"
         SIMD_FLAGS='';;
       *)
         AC_MSG_ERROR(["SIMD option ${ac_SIMD} not supported by the Intel compiler"]);;
@@ -208,14 +312,18 @@ AM_CXXFLAGS="$SIMD_FLAGS $AM_CXXFLAGS"
 AM_CFLAGS="$SIMD_FLAGS $AM_CFLAGS"
 
 case ${ac_SIMD} in
-  AVX512|AVX512MIC|KNL)
+  AVX512|KNL)
     AC_DEFINE([TEST_ZMM],[1],[compile ZMM test]);;
   *)
 	;;
 esac
 
-############### precision selection
-AC_ARG_ENABLE([precision],[AC_HELP_STRING([--enable-precision=single|double],[Select default word size of Real])],[ac_PRECISION=${enable_precision}],[ac_PRECISION=double])
+############### Precision selection
+AC_ARG_ENABLE([precision],
+              [AC_HELP_STRING([--enable-precision=single|double],
+                              [Select default word size of Real])],
+              [ac_PRECISION=${enable_precision}],[ac_PRECISION=double])
+
 case ${ac_PRECISION} in
      single)
        AC_DEFINE([GRID_DEFAULT_PRECISION_SINGLE],[1],[GRID_DEFAULT_PRECISION is SINGLE] )
@@ -223,42 +331,92 @@ case ${ac_PRECISION} in
      double)
        AC_DEFINE([GRID_DEFAULT_PRECISION_DOUBLE],[1],[GRID_DEFAULT_PRECISION is DOUBLE] )
      ;;
+     *)
+     AC_MSG_ERROR([${ac_PRECISION} unsupported --enable-precision option]);
+     ;;
 esac
 
+######################  Shared memory allocation technique under MPI3
+AC_ARG_ENABLE([shm],[AC_HELP_STRING([--enable-shm=shmget|shmopen|hugetlbfs],
+              [Select SHM allocation technique])],[ac_SHM=${enable_shm}],[ac_SHM=shmopen])
+
+case ${ac_SHM} in
+
+     shmget)
+     AC_DEFINE([GRID_MPI3_SHMGET],[1],[GRID_MPI3_SHMGET] )
+     ;;
+
+     shmopen)
+     AC_DEFINE([GRID_MPI3_SHMOPEN],[1],[GRID_MPI3_SHMOPEN] )
+     ;;
+
+     hugetlbfs)
+     AC_DEFINE([GRID_MPI3_SHMMMAP],[1],[GRID_MPI3_SHMMMAP] )
+     ;;
+
+     *)
+     AC_MSG_ERROR([${ac_SHM} unsupported --enable-shm option]);
+     ;;
+esac
+
+######################  Shared base path for SHMMMAP
+AC_ARG_ENABLE([shmpath],[AC_HELP_STRING([--enable-shmpath=path],
+              [Select SHM mmap base path for hugetlbfs])],
+	      [ac_SHMPATH=${enable_shmpath}],
+	      [ac_SHMPATH=/var/lib/hugetlbfs/pagesize-2MB/])
+AC_DEFINE_UNQUOTED([GRID_SHM_PATH],["$ac_SHMPATH"],[Path to a hugetlbfs filesystem for MMAPing])
+
 ############### communication type selection
-AC_ARG_ENABLE([comms],[AC_HELP_STRING([--enable-comms=none|mpi|mpi-auto|shmem],[Select communications])],[ac_COMMS=${enable_comms}],[ac_COMMS=none])
+AC_ARG_ENABLE([comms],[AC_HELP_STRING([--enable-comms=none|mpi|mpi-auto|mpi3|mpi3-auto|shmem],
+              [Select communications])],[ac_COMMS=${enable_comms}],[ac_COMMS=none])
 
 case ${ac_COMMS} in
      none)
-       AC_DEFINE([GRID_COMMS_NONE],[1],[GRID_COMMS_NONE] )
+        AC_DEFINE([GRID_COMMS_NONE],[1],[GRID_COMMS_NONE] )
+        comms_type='none'
      ;;
-     mpi-auto)
-       AC_DEFINE([GRID_COMMS_MPI],[1],[GRID_COMMS_MPI] )
-       LX_FIND_MPI
-       if test "x$have_CXX_mpi" = 'xno'; then AC_MSG_ERROR(["MPI not found"]); fi
-       AM_CXXFLAGS="$MPI_CXXFLAGS $AM_CXXFLAGS"
-       AM_CFLAGS="$MPI_CFLAGS $AM_CFLAGS"
-       AM_LDFLAGS="`echo $MPI_CXXLDFLAGS | sed -E 's/-l@<:@^ @:>@+//g'` $AM_LDFLAGS"
-       LIBS="`echo $MPI_CXXLDFLAGS | sed -E 's/-L@<:@^ @:>@+//g'` $LIBS"
+     mpi3*)
+        AC_DEFINE([GRID_COMMS_MPI3],[1],[GRID_COMMS_MPI3] )
+        comms_type='mpi3'
      ;;
-     mpi)
-       AC_DEFINE([GRID_COMMS_MPI],[1],[GRID_COMMS_MPI] )
+     mpit)
+        AC_DEFINE([GRID_COMMS_MPIT],[1],[GRID_COMMS_MPIT] )
+        comms_type='mpit'
+     ;;
+     mpi*)
+        AC_DEFINE([GRID_COMMS_MPI],[1],[GRID_COMMS_MPI] )
+        comms_type='mpi'
      ;;
      shmem)
-       AC_DEFINE([GRID_COMMS_SHMEM],[1],[GRID_COMMS_SHMEM] )
+        AC_DEFINE([GRID_COMMS_SHMEM],[1],[GRID_COMMS_SHMEM] )
+        comms_type='shmem'
      ;;
      *)
-     AC_MSG_ERROR([${ac_COMMS} unsupported --enable-comms option]); 
+        AC_MSG_ERROR([${ac_COMMS} unsupported --enable-comms option]);
      ;;
 esac
-AM_CONDITIONAL(BUILD_COMMS_SHMEM,[ test "X${ac_COMMS}X" == "XshmemX" ])
-AM_CONDITIONAL(BUILD_COMMS_MPI,[ test "X${ac_COMMS}X" == "XmpiX" || test "X${ac_COMMS}X" == "Xmpi-autoX" ])
-AM_CONDITIONAL(BUILD_COMMS_NONE,[ test "X${ac_COMMS}X" == "XnoneX" ])
+case ${ac_COMMS} in
+    *-auto)
+        LX_FIND_MPI
+        if test "x$have_CXX_mpi" = 'xno'; then AC_MSG_ERROR(["The configure could not find the MPI compilation flags. N.B. The -auto mode is not supported by Cray wrappers. Use the non -auto version in this case."]); fi
+        AM_CXXFLAGS="$MPI_CXXFLAGS $AM_CXXFLAGS"
+        AM_CFLAGS="$MPI_CFLAGS $AM_CFLAGS"
+        AM_LDFLAGS="`echo $MPI_CXXLDFLAGS | sed -E 's/-l@<:@^ @:>@+//g'` $AM_LDFLAGS"
+        LIBS="`echo $MPI_CXXLDFLAGS | sed -E 's/-L@<:@^ @:>@+//g'` $LIBS";;
+    *)
+        ;;
+esac
+
+AM_CONDITIONAL(BUILD_COMMS_SHMEM, [ test "${comms_type}X" == "shmemX" ])
+AM_CONDITIONAL(BUILD_COMMS_MPI,   [ test "${comms_type}X" == "mpiX" ])
+AM_CONDITIONAL(BUILD_COMMS_MPI3,  [ test "${comms_type}X" == "mpi3X" ] )
+AM_CONDITIONAL(BUILD_COMMS_MPIT,  [ test "${comms_type}X" == "mpitX" ] )
+AM_CONDITIONAL(BUILD_COMMS_NONE,  [ test "${comms_type}X" == "noneX" ])
 
 ############### RNG selection
-AC_ARG_ENABLE([rng],[AC_HELP_STRING([--enable-rng=ranlux48|mt19937],\
-	[Select Random Number Generator to be used])],\
-	[ac_RNG=${enable_rng}],[ac_RNG=ranlux48])
+AC_ARG_ENABLE([rng],[AC_HELP_STRING([--enable-rng=ranlux48|mt19937|sitmo],\
+	            [Select Random Number Generator to be used])],\
+	            [ac_RNG=${enable_rng}],[ac_RNG=sitmo])
 
 case ${ac_RNG} in
      ranlux48)
@@ -267,15 +425,19 @@ case ${ac_RNG} in
      mt19937)
       AC_DEFINE([RNG_MT19937],[1],[RNG_MT19937] )
      ;;
+     sitmo)
+      AC_DEFINE([RNG_SITMO],[1],[RNG_SITMO] )
+     ;;
      *)
-      AC_MSG_ERROR([${ac_RNG} unsupported --enable-rng option]); 
+      AC_MSG_ERROR([${ac_RNG} unsupported --enable-rng option]);
      ;;
 esac
 
-############### timer option
+############### Timer option
 AC_ARG_ENABLE([timers],[AC_HELP_STRING([--enable-timers],\
-	[Enable system dependent high res timers])],\
-	[ac_TIMERS=${enable_timers}],[ac_TIMERS=yes])
+	            [Enable system dependent high res timers])],\
+	            [ac_TIMERS=${enable_timers}],[ac_TIMERS=yes])
+
 case ${ac_TIMERS} in
      yes)
       AC_DEFINE([TIMERS_ON],[1],[TIMERS_ON] )
@@ -284,37 +446,100 @@ case ${ac_TIMERS} in
       AC_DEFINE([TIMERS_OFF],[1],[TIMERS_OFF] )
      ;;
      *)
-      AC_MSG_ERROR([${ac_TIMERS} unsupported --enable-timers option]); 
+      AC_MSG_ERROR([${ac_TIMERS} unsupported --enable-timers option]);
      ;;
 esac
 
 ############### Chroma regression test
-AC_ARG_ENABLE([chroma],[AC_HELP_STRING([--enable-chroma],[Expect chroma compiled under c++11 ])],ac_CHROMA=yes,ac_CHROMA=no)
+AC_ARG_ENABLE([chroma],[AC_HELP_STRING([--enable-chroma],
+              [Expect chroma compiled under c++11 ])],ac_CHROMA=yes,ac_CHROMA=no)
+
 case ${ac_CHROMA} in
      yes|no)
      ;;
      *)
-       AC_MSG_ERROR([${ac_CHROMA} unsupported --enable-chroma option]); 
+       AC_MSG_ERROR([${ac_CHROMA} unsupported --enable-chroma option]);
      ;;
 esac
+
 AM_CONDITIONAL(BUILD_CHROMA_REGRESSION,[ test "X${ac_CHROMA}X" == "XyesX" ])
 
 ############### Doxygen
-AC_PROG_DOXYGEN
-
-if test -n "$DOXYGEN"
-then
-AC_CONFIG_FILES([docs/doxy.cfg])
-fi
+DX_DOXYGEN_FEATURE([OFF])
+DX_DOT_FEATURE([OFF])
+DX_HTML_FEATURE([ON])
+DX_CHM_FEATURE([OFF])
+DX_CHI_FEATURE([OFF])
+DX_MAN_FEATURE([OFF])
+DX_RTF_FEATURE([OFF])
+DX_XML_FEATURE([OFF])
+DX_PDF_FEATURE([OFF])
+DX_PS_FEATURE([OFF])
+DX_INIT_DOXYGEN([$PACKAGE_NAME], [doxygen.cfg])
 
 ############### Ouput
 cwd=`pwd -P`; cd ${srcdir}; abs_srcdir=`pwd -P`; cd ${cwd}
+GRID_CXXFLAGS="$AM_CXXFLAGS $CXXFLAGS"
+GRID_LDFLAGS="$AM_LDFLAGS $LDFLAGS"
+GRID_LIBS=$LIBS
+GRID_SHORT_SHA=`git rev-parse --short HEAD`
+GRID_SHA=`git rev-parse HEAD`
+GRID_BRANCH=`git rev-parse --abbrev-ref HEAD`
 AM_CXXFLAGS="-I${abs_srcdir}/include $AM_CXXFLAGS"
 AM_CFLAGS="-I${abs_srcdir}/include $AM_CFLAGS"
 AM_LDFLAGS="-L${cwd}/lib $AM_LDFLAGS"
 AC_SUBST([AM_CFLAGS])
 AC_SUBST([AM_CXXFLAGS])
 AC_SUBST([AM_LDFLAGS])
+AC_SUBST([GRID_CXXFLAGS])
+AC_SUBST([GRID_LDFLAGS])
+AC_SUBST([GRID_LIBS])
+AC_SUBST([GRID_SHA])
+AC_SUBST([GRID_BRANCH])
+
+git_commit=`cd $srcdir && ./scripts/configure.commit`
+
+echo "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+Summary of configuration for $PACKAGE v$VERSION
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+----- GIT VERSION -------------------------------------
+$git_commit
+----- PLATFORM ----------------------------------------
+architecture (build)        : $build_cpu
+os (build)                  : $build_os
+architecture (target)       : $target_cpu
+os (target)                 : $target_os
+compiler vendor             : ${ax_cv_cxx_compiler_vendor}
+compiler version            : ${ax_cv_gxx_version}
+----- BUILD OPTIONS -----------------------------------
+SIMD                        : ${ac_SIMD}${SIMD_GEN_WIDTH_MSG}
+Threading                   : ${ac_openmp}
+Communications type         : ${comms_type}
+Shared memory allocator     : ${ac_SHM}
+Shared memory mmap path     : ${ac_SHMPATH}
+Default precision           : ${ac_PRECISION}
+Software FP16 conversion    : ${ac_SFW_FP16}
+RNG choice                  : ${ac_RNG}
+GMP                         : `if test "x$have_gmp" = xtrue; then echo yes; else echo no; fi`
+LAPACK                      : ${ac_LAPACK}
+FFTW                        : `if test "x$have_fftw" = xtrue; then echo yes; else echo no; fi`
+LIME (ILDG support)         : `if test "x$have_lime" = xtrue; then echo yes; else echo no; fi`
+HDF5                        : `if test "x$have_hdf5" = xtrue; then echo yes; else echo no; fi`
+build DOXYGEN documentation : `if test "$DX_FLAG_doc" = '1'; then echo yes; else echo no; fi`
+----- BUILD FLAGS -------------------------------------
+CXXFLAGS:
+`echo ${AM_CXXFLAGS} ${CXXFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
+LDFLAGS:
+`echo ${AM_LDFLAGS} ${LDFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
+LIBS:
+`echo ${LIBS} | tr ' ' '\n' | sed 's/^-/    -/g'`
+-------------------------------------------------------" > grid.configure.summary
+
+GRID_SUMMARY="`cat grid.configure.summary`"
+AM_SUBST_NOTMAKE([GRID_SUMMARY])
+AC_SUBST([GRID_SUMMARY])
+
+AC_CONFIG_FILES([grid-config], [chmod +x grid-config])
 AC_CONFIG_FILES(Makefile)
 AC_CONFIG_FILES(lib/Makefile)
 AC_CONFIG_FILES(tests/Makefile)
@@ -322,41 +547,18 @@ AC_CONFIG_FILES(tests/IO/Makefile)
 AC_CONFIG_FILES(tests/core/Makefile)
 AC_CONFIG_FILES(tests/debug/Makefile)
 AC_CONFIG_FILES(tests/forces/Makefile)
+AC_CONFIG_FILES(tests/hadrons/Makefile)
 AC_CONFIG_FILES(tests/hmc/Makefile)
 AC_CONFIG_FILES(tests/solver/Makefile)
+AC_CONFIG_FILES(tests/smearing/Makefile)
 AC_CONFIG_FILES(tests/qdpxx/Makefile)
+AC_CONFIG_FILES(tests/testu01/Makefile)
 AC_CONFIG_FILES(benchmarks/Makefile)
+AC_CONFIG_FILES(extras/Makefile)
+AC_CONFIG_FILES(extras/Hadrons/Makefile)
 AC_OUTPUT
 
-echo "
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-Summary of configuration for $PACKAGE v$VERSION
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+echo ""
+cat grid.configure.summary
+echo ""
 
------ PLATFORM ----------------------------------------
-- architecture (build)          : $build_cpu
-- os (build)                    : $build_os
-- architecture (target)         : $target_cpu
-- os (target)                   : $target_os
-- compiler vendor               : ${ax_cv_cxx_compiler_vendor}
-- compiler version              : ${ax_cv_gxx_version}
------ BUILD OPTIONS -----------------------------------
-- SIMD                          : ${ac_SIMD}
-- Threading                     : ${ac_openmp} 
-- Communications type           : ${ac_COMMS}
-- Default precision             : ${ac_PRECISION}
-- RNG choice                    : ${ac_RNG} 
-- GMP                           : `if test "x$have_gmp" = xtrue; then echo yes; else echo no; fi`
-- LAPACK                        : ${ac_LAPACK}
-- FFTW                          : `if test "x$have_fftw" = xtrue; then echo yes; else echo no; fi`
-- build DOXYGEN documentation   : `if test "x$enable_doc" = xyes; then echo yes; else echo no; fi`
-- graphs and diagrams           : `if test "x$enable_dot" = xyes; then echo yes; else echo no; fi`
------ BUILD FLAGS -------------------------------------
-- CXXFLAGS:
-`echo ${AM_CXXFLAGS} ${CXXFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
-- LDFLAGS:
-`echo ${AM_LDFLAGS} ${LDFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
-- LIBS:
-`echo ${LIBS} | tr ' ' '\n' | sed 's/^-/    -/g'`
--------------------------------------------------------
-"

doxygen.inc

@@ -0,0 +1,184 @@
+# Copyright (C) 2004 Oren Ben-Kiki
+# This file is distributed under the same terms as the Automake macro files.
+
+# Generate automatic documentation using Doxygen. Goals and variables values
+# are controlled by the various DX_COND_??? conditionals set by autoconf.
+#
+# The provided goals are:
+# doxygen-doc: Generate all doxygen documentation.
+# doxygen-run: Run doxygen, which will generate some of the documentation
+#              (HTML, CHM, CHI, MAN, RTF, XML) but will not do the post
+#              processing required for the rest of it (PS, PDF, and some MAN).
+# doxygen-man: Rename some doxygen generated man pages.
+# doxygen-ps: Generate doxygen PostScript documentation.
+# doxygen-pdf: Generate doxygen PDF documentation.
+#
+# Note that by default these are not integrated into the automake goals. If
+# doxygen is used to generate man pages, you can achieve this integration by
+# setting man3_MANS to the list of man pages generated and then adding the
+# dependency:
+#
+#   $(man3_MANS): doxygen-doc
+#
+# This will cause make to run doxygen and generate all the documentation.
+#
+# The following variable is intended for use in Makefile.am:
+#
+# DX_CLEANFILES = everything to clean.
+#
+# This is usually added to MOSTLYCLEANFILES.
+
+## --------------------------------- ##
+## Format-independent Doxygen rules. ##
+## --------------------------------- ##
+
+if DX_COND_doc
+
+## ------------------------------- ##
+## Rules specific for HTML output. ##
+## ------------------------------- ##
+
+if DX_COND_html
+
+DX_CLEAN_HTML = @DX_DOCDIR@/html
+
+endif DX_COND_html
+
+## ------------------------------ ##
+## Rules specific for CHM output. ##
+## ------------------------------ ##
+
+if DX_COND_chm
+
+DX_CLEAN_CHM = @DX_DOCDIR@/chm
+
+if DX_COND_chi
+
+DX_CLEAN_CHI = @DX_DOCDIR@/@PACKAGE@.chi
+
+endif DX_COND_chi
+
+endif DX_COND_chm
+
+## ------------------------------ ##
+## Rules specific for MAN output. ##
+## ------------------------------ ##
+
+if DX_COND_man
+
+DX_CLEAN_MAN = @DX_DOCDIR@/man
+
+endif DX_COND_man
+
+## ------------------------------ ##
+## Rules specific for RTF output. ##
+## ------------------------------ ##
+
+if DX_COND_rtf
+
+DX_CLEAN_RTF = @DX_DOCDIR@/rtf
+
+endif DX_COND_rtf
+
+## ------------------------------ ##
+## Rules specific for XML output. ##
+## ------------------------------ ##
+
+if DX_COND_xml
+
+DX_CLEAN_XML = @DX_DOCDIR@/xml
+
+endif DX_COND_xml
+
+## ----------------------------- ##
+## Rules specific for PS output. ##
+## ----------------------------- ##
+
+if DX_COND_ps
+
+DX_CLEAN_PS = @DX_DOCDIR@/@PACKAGE@.ps
+
+DX_PS_GOAL = doxygen-ps
+
+doxygen-ps: @DX_DOCDIR@/@PACKAGE@.ps
+
+@DX_DOCDIR@/@PACKAGE@.ps: @DX_DOCDIR@/@PACKAGE@.tag
+	cd @DX_DOCDIR@/latex; \
+	rm -f *.aux *.toc *.idx *.ind *.ilg *.log *.out; \
+	$(DX_LATEX) refman.tex; \
+	$(MAKEINDEX_PATH) refman.idx; \
+	$(DX_LATEX) refman.tex; \
+	countdown=5; \
+	while $(DX_EGREP) 'Rerun (LaTeX|to get cross-references right)' \
+	                  refman.log > /dev/null 2>&1 \
+	   && test $$countdown -gt 0; do \
+	    $(DX_LATEX) refman.tex; \
+	    countdown=`expr $$countdown - 1`; \
+	done; \
+	$(DX_DVIPS) -o ../@PACKAGE@.ps refman.dvi
+
+endif DX_COND_ps
+
+## ------------------------------ ##
+## Rules specific for PDF output. ##
+## ------------------------------ ##
+
+if DX_COND_pdf
+
+DX_CLEAN_PDF = @DX_DOCDIR@/@PACKAGE@.pdf
+
+DX_PDF_GOAL = doxygen-pdf
+
+doxygen-pdf: @DX_DOCDIR@/@PACKAGE@.pdf
+
+@DX_DOCDIR@/@PACKAGE@.pdf: @DX_DOCDIR@/@PACKAGE@.tag
+	cd @DX_DOCDIR@/latex; \
+	rm -f *.aux *.toc *.idx *.ind *.ilg *.log *.out; \
+	$(DX_PDFLATEX) refman.tex; \
+	$(DX_MAKEINDEX) refman.idx; \
+	$(DX_PDFLATEX) refman.tex; \
+	countdown=5; \
+	while $(DX_EGREP) 'Rerun (LaTeX|to get cross-references right)' \
+	                  refman.log > /dev/null 2>&1 \
+	   && test $$countdown -gt 0; do \
+	    $(DX_PDFLATEX) refman.tex; \
+	    countdown=`expr $$countdown - 1`; \
+	done; \
+	mv refman.pdf ../@PACKAGE@.pdf
+
+endif DX_COND_pdf
+
+## ------------------------------------------------- ##
+## Rules specific for LaTeX (shared for PS and PDF). ##
+## ------------------------------------------------- ##
+
+if DX_COND_latex
+
+DX_CLEAN_LATEX = @DX_DOCDIR@/latex
+
+endif DX_COND_latex
+
+.INTERMEDIATE: doxygen-run $(DX_PS_GOAL) $(DX_PDF_GOAL)
+
+doxygen-run: @DX_DOCDIR@/@PACKAGE@.tag
+
+doxygen-doc: doxygen-run $(DX_PS_GOAL) $(DX_PDF_GOAL)
+
+@DX_DOCDIR@/@PACKAGE@.tag: $(DX_CONFIG) $(pkginclude_HEADERS)
+	rm -rf @DX_DOCDIR@
+	$(DX_ENV) $(DX_DOXYGEN) $(srcdir)/$(DX_CONFIG)
+
+DX_CLEANFILES = \
+    @DX_DOCDIR@/@PACKAGE@.tag \
+    -r \
+    $(DX_CLEAN_HTML) \
+    $(DX_CLEAN_CHM) \
+    $(DX_CLEAN_CHI) \
+    $(DX_CLEAN_MAN) \
+    $(DX_CLEAN_RTF) \
+    $(DX_CLEAN_XML) \
+    $(DX_CLEAN_PS) \
+    $(DX_CLEAN_PDF) \
+    $(DX_CLEAN_LATEX)
+
+endif DX_COND_doc

extras/Hadrons/Application.cc

@@ -0,0 +1,318 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Application.cc
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Hadrons/Application.hpp>
+#include <Grid/Hadrons/GeneticScheduler.hpp>
+
+using namespace Grid;
+using namespace QCD;
+using namespace Hadrons;
+
+#define BIG_SEP "==============="
+#define SEP     "---------------"
+
+/******************************************************************************
+ *                       Application implementation                           *
+ ******************************************************************************/
+// constructors ////////////////////////////////////////////////////////////////
+Application::Application(void)
+{
+    LOG(Message) << "Modules available:" << std::endl;
+    auto list = ModuleFactory::getInstance().getBuilderList();
+    for (auto &m: list)
+    {
+        LOG(Message) << "  " << m << std::endl;
+    }
+    auto dim = GridDefaultLatt(), mpi = GridDefaultMpi(), loc(dim);
+    locVol_ = 1;
+    for (unsigned int d = 0; d < dim.size(); ++d)
+    {
+        loc[d]  /= mpi[d];
+        locVol_ *= loc[d];
+    }
+    LOG(Message) << "Global lattice: " << dim << std::endl;
+    LOG(Message) << "MPI partition : " << mpi << std::endl;
+    LOG(Message) << "Local lattice : " << loc << std::endl;
+}
+
+Application::Application(const Application::GlobalPar &par)
+: Application()
+{
+    setPar(par);
+}
+
+Application::Application(const std::string parameterFileName)
+: Application()
+{
+    parameterFileName_ = parameterFileName;
+}
+
+// environment shortcut ////////////////////////////////////////////////////////
+Environment & Application::env(void) const
+{
+    return Environment::getInstance();
+}
+
+// access //////////////////////////////////////////////////////////////////////
+void Application::setPar(const Application::GlobalPar &par)
+{
+    par_ = par;
+    env().setSeed(strToVec<int>(par_.seed));
+}
+
+const Application::GlobalPar & Application::getPar(void)
+{
+    return par_;
+}
+
+// execute /////////////////////////////////////////////////////////////////////
+void Application::run(void)
+{
+    if (!parameterFileName_.empty() and (env().getNModule() == 0))
+    {
+        parseParameterFile(parameterFileName_);
+    }
+    if (!scheduled_)
+    {
+        schedule();
+    }
+    printSchedule();
+    configLoop();
+}
+
+// parse parameter file ////////////////////////////////////////////////////////
+class ObjectId: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(ObjectId,
+                                    std::string, name,
+                                    std::string, type);
+};
+
+void Application::parseParameterFile(const std::string parameterFileName)
+{
+    XmlReader reader(parameterFileName);
+    GlobalPar par;
+    ObjectId  id;
+    
+    LOG(Message) << "Building application from '" << parameterFileName << "'..." << std::endl;
+    read(reader, "parameters", par);
+    setPar(par);
+    push(reader, "modules");
+    push(reader, "module");
+    do
+    {
+        read(reader, "id", id);
+        env().createModule(id.name, id.type, reader);
+    } while (reader.nextElement("module"));
+    pop(reader);
+    pop(reader);
+}
+
+void Application::saveParameterFile(const std::string parameterFileName)
+{
+    XmlWriter          writer(parameterFileName);
+    ObjectId           id;
+    const unsigned int nMod = env().getNModule();
+    
+    LOG(Message) << "Saving application to '" << parameterFileName << "'..." << std::endl;
+    write(writer, "parameters", getPar());
+    push(writer, "modules");
+    for (unsigned int i = 0; i < nMod; ++i)
+    {
+        push(writer, "module");
+        id.name = env().getModuleName(i);
+        id.type = env().getModule(i)->getRegisteredName();
+        write(writer, "id", id);
+        env().getModule(i)->saveParameters(writer, "options");
+        pop(writer);
+    }
+    pop(writer);
+    pop(writer);
+}
+
+// schedule computation ////////////////////////////////////////////////////////
+#define MEM_MSG(size)\
+sizeString((size)*locVol_) << " (" << sizeString(size)  << "/site)"
+
+#define DEFINE_MEMPEAK \
+GeneticScheduler<unsigned int>::ObjFunc memPeak = \
+[this](const std::vector<unsigned int> &program)\
+{\
+    unsigned int memPeak;\
+    bool         msg;\
+    \
+    msg = HadronsLogMessage.isActive();\
+    HadronsLogMessage.Active(false);\
+    env().dryRun(true);\
+    memPeak = env().executeProgram(program);\
+    env().dryRun(false);\
+    env().freeAll();\
+    HadronsLogMessage.Active(true);\
+    \
+    return memPeak;\
+}
+
+void Application::schedule(void)
+{
+    DEFINE_MEMPEAK;
+    
+    // build module dependency graph
+    LOG(Message) << "Building module graph..." << std::endl;
+    auto graph = env().makeModuleGraph();
+    auto con = graph.getConnectedComponents();
+    
+    // constrained topological sort using a genetic algorithm
+    LOG(Message) << "Scheduling computation..." << std::endl;
+    LOG(Message) << "               #module= " << graph.size() << std::endl;
+    LOG(Message) << "       population size= " << par_.genetic.popSize << std::endl;
+    LOG(Message) << "       max. generation= " << par_.genetic.maxGen << std::endl;
+    LOG(Message) << "  max. cst. generation= " << par_.genetic.maxCstGen << std::endl;
+    LOG(Message) << "         mutation rate= " << par_.genetic.mutationRate << std::endl;
+    
+    unsigned int                               k = 0, gen, prevPeak, nCstPeak = 0;
+    std::random_device                         rd;
+    GeneticScheduler<unsigned int>::Parameters par;
+    
+    par.popSize      = par_.genetic.popSize;
+    par.mutationRate = par_.genetic.mutationRate;
+    par.seed         = rd();
+    memPeak_         = 0;
+    CartesianCommunicator::BroadcastWorld(0, &(par.seed), sizeof(par.seed));
+    for (unsigned int i = 0; i < con.size(); ++i)
+    {
+        GeneticScheduler<unsigned int> scheduler(con[i], memPeak, par);
+        
+        gen = 0;
+        do
+        {
+            LOG(Debug) << "Generation " << gen << ":" << std::endl;
+            scheduler.nextGeneration();
+            if (gen != 0)
+            {
+                if (prevPeak == scheduler.getMinValue())
+                {
+                    nCstPeak++;
+                }
+                else
+                {
+                    nCstPeak = 0;
+                }
+            }
+            
+            prevPeak = scheduler.getMinValue();
+            if (gen % 10 == 0)
+            {
+                LOG(Iterative) << "Generation " << gen << ": "
+                               << MEM_MSG(scheduler.getMinValue()) << std::endl;
+            }
+            
+            gen++;
+        } while ((gen < par_.genetic.maxGen)
+                 and (nCstPeak < par_.genetic.maxCstGen));
+        auto &t = scheduler.getMinSchedule();
+        if (scheduler.getMinValue() > memPeak_)
+        {
+            memPeak_ = scheduler.getMinValue();
+        }
+        for (unsigned int j = 0; j < t.size(); ++j)
+        {
+            program_.push_back(t[j]);
+        }
+    }
+    scheduled_ = true;
+}
+
+void Application::saveSchedule(const std::string filename)
+{
+    TextWriter               writer(filename);
+    std::vector<std::string> program;
+    
+    if (!scheduled_)
+    {
+        HADRON_ERROR("Computation not scheduled");
+    }
+    LOG(Message) << "Saving current schedule to '" << filename << "'..."
+                 << std::endl;
+    for (auto address: program_)
+    {
+        program.push_back(env().getModuleName(address));
+    }
+    write(writer, "schedule", program);
+}
+
+void Application::loadSchedule(const std::string filename)
+{
+    DEFINE_MEMPEAK;
+    
+    TextReader               reader(filename);
+    std::vector<std::string> program;
+    
+    LOG(Message) << "Loading schedule from '" << filename << "'..."
+                 << std::endl;
+    read(reader, "schedule", program);
+    program_.clear();
+    for (auto &name: program)
+    {
+        program_.push_back(env().getModuleAddress(name));
+    }
+    scheduled_ = true;
+    memPeak_   = memPeak(program_);
+}
+
+void Application::printSchedule(void)
+{
+    if (!scheduled_)
+    {
+        HADRON_ERROR("Computation not scheduled");
+    }
+    LOG(Message) << "Schedule (memory peak: " << MEM_MSG(memPeak_) << "):"
+                 << std::endl;
+    for (unsigned int i = 0; i < program_.size(); ++i)
+    {
+        LOG(Message) << std::setw(4) << i + 1 << ": "
+                     << env().getModuleName(program_[i]) << std::endl;
+    }
+}
+
+// loop on configurations //////////////////////////////////////////////////////
+void Application::configLoop(void)
+{
+    auto range = par_.trajCounter;
+    
+    for (unsigned int t = range.start; t < range.end; t += range.step)
+    {
+        LOG(Message) << BIG_SEP << " Starting measurement for trajectory " << t
+                     << " " << BIG_SEP << std::endl;
+        env().setTrajectory(t);
+        env().executeProgram(program_);
+    }
+    LOG(Message) << BIG_SEP << " End of measurement " << BIG_SEP << std::endl;
+    env().freeAll();
+}

extras/Hadrons/Application.hpp

@@ -0,0 +1,132 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Application.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_Application_hpp_
+#define Hadrons_Application_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Environment.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+#include <Grid/Hadrons/Modules.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         Main program manager                               *
+ ******************************************************************************/
+class Application
+{
+public:
+    class TrajRange: Serializable
+    {
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(TrajRange,
+                                        unsigned int, start,
+                                        unsigned int, end,
+                                        unsigned int, step);
+    };
+    class GeneticPar: Serializable
+    {
+    public:
+        GeneticPar(void):
+            popSize{20}, maxGen{1000}, maxCstGen{100}, mutationRate{.1} {};
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(GeneticPar,
+                                        unsigned int, popSize,
+                                        unsigned int, maxGen,
+                                        unsigned int, maxCstGen,
+                                        double      , mutationRate);
+    };
+    class GlobalPar: Serializable
+    {
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(GlobalPar,
+                                        TrajRange,   trajCounter,
+                                        GeneticPar,  genetic,
+                                        std::string, seed);
+    };
+public:
+    // constructors
+    Application(void);
+    Application(const GlobalPar &par);
+    Application(const std::string parameterFileName);
+    // destructor
+    virtual ~Application(void) = default;
+    // access
+    void              setPar(const GlobalPar &par);
+    const GlobalPar & getPar(void);
+    // module creation
+    template <typename M>
+    void createModule(const std::string name);
+    template <typename M>
+    void createModule(const std::string name, const typename M::Par &par);
+    // execute
+    void run(void);
+    // XML parameter file I/O
+    void parseParameterFile(const std::string parameterFileName);
+    void saveParameterFile(const std::string parameterFileName);
+    // schedule computation
+    void schedule(void);
+    void saveSchedule(const std::string filename);
+    void loadSchedule(const std::string filename);
+    void printSchedule(void);
+    // loop on configurations
+    void configLoop(void);
+private:
+    // environment shortcut
+    Environment & env(void) const;
+private:
+    long unsigned int         locVol_;
+    std::string               parameterFileName_{""};
+    GlobalPar                 par_;
+    std::vector<unsigned int> program_;
+    Environment::Size         memPeak_;
+    bool                      scheduled_{false};
+};
+
+/******************************************************************************
+ *                     Application template implementation                    *
+ ******************************************************************************/
+// module creation /////////////////////////////////////////////////////////////
+template <typename M>
+void Application::createModule(const std::string name)
+{
+    env().createModule<M>(name);
+}
+
+template <typename M>
+void Application::createModule(const std::string name,
+                               const typename M::Par &par)
+{
+    env().createModule<M>(name, par);
+}
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_Application_hpp_

extras/Hadrons/Environment.cc

@@ -0,0 +1,793 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Environment.cc
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Hadrons/Environment.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+using namespace Grid;
+using namespace QCD;
+using namespace Hadrons;
+
+/******************************************************************************
+ *                       Environment implementation                           *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+Environment::Environment(void)
+{
+    dim_ = GridDefaultLatt();
+    nd_  = dim_.size();
+    grid4d_.reset(SpaceTimeGrid::makeFourDimGrid(
+        dim_, GridDefaultSimd(nd_, vComplex::Nsimd()),
+        GridDefaultMpi()));
+    gridRb4d_.reset(SpaceTimeGrid::makeFourDimRedBlackGrid(grid4d_.get()));
+    auto loc = getGrid()->LocalDimensions();
+    locVol_ = 1;
+    for (unsigned int d = 0; d < loc.size(); ++d)
+    {
+        locVol_ *= loc[d];
+    }
+    rng4d_.reset(new GridParallelRNG(grid4d_.get()));
+}
+
+// dry run /////////////////////////////////////////////////////////////////////
+void Environment::dryRun(const bool isDry)
+{
+    dryRun_ = isDry;
+}
+
+bool Environment::isDryRun(void) const
+{
+    return dryRun_;
+}
+
+// trajectory number ///////////////////////////////////////////////////////////
+void Environment::setTrajectory(const unsigned int traj)
+{
+    traj_ = traj;
+}
+
+unsigned int Environment::getTrajectory(void) const
+{
+    return traj_;
+}
+
+// grids ///////////////////////////////////////////////////////////////////////
+void Environment::createGrid(const unsigned int Ls)
+{
+    if (grid5d_.find(Ls) == grid5d_.end())
+    {
+        auto g = getGrid();
+        
+        grid5d_[Ls].reset(SpaceTimeGrid::makeFiveDimGrid(Ls, g));
+        gridRb5d_[Ls].reset(SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls, g));
+    }
+}
+
+GridCartesian * Environment::getGrid(const unsigned int Ls) const
+{
+    try
+    {
+        if (Ls == 1)
+        {
+            return grid4d_.get();
+        }
+        else
+        {
+            return grid5d_.at(Ls).get();
+        }
+    }
+    catch(std::out_of_range &)
+    {
+        HADRON_ERROR("no grid with Ls= " << Ls);
+    }
+}
+
+GridRedBlackCartesian * Environment::getRbGrid(const unsigned int Ls) const
+{
+    try
+    {
+        if (Ls == 1)
+        {
+            return gridRb4d_.get();
+        }
+        else
+        {
+            return gridRb5d_.at(Ls).get();
+        }
+    }
+    catch(std::out_of_range &)
+    {
+        HADRON_ERROR("no red-black 5D grid with Ls= " << Ls);
+    }
+}
+
+unsigned int Environment::getNd(void) const
+{
+    return nd_;
+}
+
+std::vector<int> Environment::getDim(void) const
+{
+    return dim_;
+}
+
+int Environment::getDim(const unsigned int mu) const
+{
+    return dim_[mu];
+}
+
+// random number generator /////////////////////////////////////////////////////
+void Environment::setSeed(const std::vector<int> &seed)
+{
+    rng4d_->SeedFixedIntegers(seed);
+}
+
+GridParallelRNG * Environment::get4dRng(void) const
+{
+    return rng4d_.get();
+}
+
+// module management ///////////////////////////////////////////////////////////
+void Environment::pushModule(Environment::ModPt &pt)
+{
+    std::string name = pt->getName();
+    
+    if (!hasModule(name))
+    {
+        std::vector<unsigned int> inputAddress;
+        unsigned int              address;
+        ModuleInfo                m;
+        
+        m.data = std::move(pt);
+        m.type = typeIdPt(*m.data.get());
+        m.name = name;
+        auto input  = m.data->getInput();
+        for (auto &in: input)
+        {
+            if (!hasObject(in))
+            {
+                addObject(in , -1);
+            }
+            m.input.push_back(objectAddress_[in]);
+        }
+        auto output = m.data->getOutput();
+        module_.push_back(std::move(m));
+        address              = static_cast<unsigned int>(module_.size() - 1);
+        moduleAddress_[name] = address;
+        for (auto &out: output)
+        {
+            if (!hasObject(out))
+            {
+                addObject(out, address);
+            }
+            else
+            {
+                if (object_[objectAddress_[out]].module < 0)
+                {
+                    object_[objectAddress_[out]].module = address;
+                }
+                else
+                {
+                    HADRON_ERROR("object '" + out
+                                 + "' is already produced by module '"
+                                 + module_[object_[getObjectAddress(out)].module].name
+                                 + "' (while pushing module '" + name + "')");
+                }
+            }
+        }
+    }
+    else
+    {
+        HADRON_ERROR("module '" + name + "' already exists");
+    }
+}
+
+unsigned int Environment::getNModule(void) const
+{
+    return module_.size();
+}
+
+void Environment::createModule(const std::string name, const std::string type,
+                               XmlReader &reader)
+{
+    auto &factory = ModuleFactory::getInstance();
+    auto pt       = factory.create(type, name);
+    
+    pt->parseParameters(reader, "options");
+    pushModule(pt);
+}
+
+ModuleBase * Environment::getModule(const unsigned int address) const
+{
+    if (hasModule(address))
+    {
+        return module_[address].data.get();
+    }
+    else
+    {
+        HADRON_ERROR("no module with address " + std::to_string(address));
+    }
+}
+
+ModuleBase * Environment::getModule(const std::string name) const
+{
+    return getModule(getModuleAddress(name));
+}
+
+unsigned int Environment::getModuleAddress(const std::string name) const
+{
+    if (hasModule(name))
+    {
+        return moduleAddress_.at(name);
+    }
+    else
+    {
+        HADRON_ERROR("no module with name '" + name + "'");
+    }
+}
+
+std::string Environment::getModuleName(const unsigned int address) const
+{
+    if (hasModule(address))
+    {
+        return module_[address].name;
+    }
+    else
+    {
+        HADRON_ERROR("no module with address " + std::to_string(address));
+    }
+}
+
+std::string Environment::getModuleType(const unsigned int address) const
+{
+    if (hasModule(address))
+    {
+        return typeName(module_[address].type);
+    }
+    else
+    {
+        HADRON_ERROR("no module with address " + std::to_string(address));
+    }
+}
+
+std::string Environment::getModuleType(const std::string name) const
+{
+    return getModuleType(getModuleAddress(name));
+}
+
+std::string Environment::getModuleNamespace(const unsigned int address) const
+{
+    std::string type = getModuleType(address), ns;
+    
+    auto pos2 = type.rfind("::");
+    auto pos1 = type.rfind("::", pos2 - 2);
+    
+    return type.substr(pos1 + 2, pos2 - pos1 - 2);
+}
+
+std::string Environment::getModuleNamespace(const std::string name) const
+{
+    return getModuleNamespace(getModuleAddress(name));
+}
+
+bool Environment::hasModule(const unsigned int address) const
+{
+    return (address < module_.size());
+}
+
+bool Environment::hasModule(const std::string name) const
+{
+    return (moduleAddress_.find(name) != moduleAddress_.end());
+}
+
+Graph<unsigned int> Environment::makeModuleGraph(void) const
+{
+    Graph<unsigned int> moduleGraph;
+    
+    for (unsigned int i = 0; i < module_.size(); ++i)
+    {
+        moduleGraph.addVertex(i);
+        for (auto &j: module_[i].input)
+        {
+            moduleGraph.addEdge(object_[j].module, i);
+        }
+    }
+    
+    return moduleGraph;
+}
+
+#define BIG_SEP "==============="
+#define SEP     "---------------"
+#define MEM_MSG(size)\
+sizeString((size)*locVol_) << " (" << sizeString(size)  << "/site)"
+
+Environment::Size
+Environment::executeProgram(const std::vector<unsigned int> &p)
+{
+    Size                                memPeak = 0, sizeBefore, sizeAfter;
+    std::vector<std::set<unsigned int>> freeProg;
+    bool                                continueCollect, nothingFreed;
+    
+    // build garbage collection schedule
+    freeProg.resize(p.size());
+    for (unsigned int i = 0; i < object_.size(); ++i)
+    {
+        auto pred = [i, this](const unsigned int j)
+        {
+            auto &in = module_[j].input;
+            auto it  = std::find(in.begin(), in.end(), i);
+            
+            return (it != in.end()) or (j == object_[i].module);
+        };
+        auto it = std::find_if(p.rbegin(), p.rend(), pred);
+        if (it != p.rend())
+        {
+            freeProg[p.rend() - it - 1].insert(i);
+        }
+    }
+    
+    // program execution
+    for (unsigned int i = 0; i < p.size(); ++i)
+    {
+        // execute module
+        if (!isDryRun())
+        {
+            LOG(Message) << SEP << " Measurement step " << i+1 << "/"
+                         << p.size() << " (module '" << module_[p[i]].name
+                         << "') " << SEP << std::endl;
+        }
+        (*module_[p[i]].data)();
+        sizeBefore = getTotalSize();
+        // print used memory after execution
+        if (!isDryRun())
+        {
+            LOG(Message) << "Allocated objects: " << MEM_MSG(sizeBefore)
+                         << std::endl;
+        }
+        if (sizeBefore > memPeak)
+        {
+            memPeak = sizeBefore;
+        }
+        // garbage collection for step i
+        if (!isDryRun())
+        {
+            LOG(Message) << "Garbage collection..." << std::endl;
+        }
+        nothingFreed = true;
+        do
+        {
+            continueCollect = false;
+            auto toFree = freeProg[i];
+            for (auto &j: toFree)
+            {
+                // continue garbage collection while there are still
+                // objects without owners
+                continueCollect = continueCollect or !hasOwners(j);
+                if(freeObject(j))
+                {
+                    // if an object has been freed, remove it from
+                    // the garbage collection schedule
+                    freeProg[i].erase(j);
+                    nothingFreed = false;
+                }
+            }
+        } while (continueCollect);
+        // any remaining objects in step i garbage collection schedule
+        // is scheduled for step i + 1
+        if (i + 1 < p.size())
+        {
+            for (auto &j: freeProg[i])
+            {
+                freeProg[i + 1].insert(j);
+            }
+        }
+        // print used memory after garbage collection if necessary
+        if (!isDryRun())
+        {
+            sizeAfter = getTotalSize();
+            if (sizeBefore != sizeAfter)
+            {
+                LOG(Message) << "Allocated objects: " << MEM_MSG(sizeAfter)
+                             << std::endl;
+            }
+            else
+            {
+                LOG(Message) << "Nothing to free" << std::endl;
+            }
+        }
+    }
+    
+    return memPeak;
+}
+
+Environment::Size Environment::executeProgram(const std::vector<std::string> &p)
+{
+    std::vector<unsigned int> pAddress;
+    
+    for (auto &n: p)
+    {
+        pAddress.push_back(getModuleAddress(n));
+    }
+    
+    return executeProgram(pAddress);
+}
+
+// general memory management ///////////////////////////////////////////////////
+void Environment::addObject(const std::string name, const int moduleAddress)
+{
+    if (!hasObject(name))
+    {
+        ObjInfo info;
+        
+        info.name   = name;
+        info.module = moduleAddress;
+        object_.push_back(std::move(info));
+        objectAddress_[name] = static_cast<unsigned int>(object_.size() - 1);
+    }
+    else
+    {
+        HADRON_ERROR("object '" + name + "' already exists");
+    }
+}
+
+void Environment::registerObject(const unsigned int address,
+                                 const unsigned int size, const unsigned int Ls)
+{
+    if (!hasRegisteredObject(address))
+    {
+        if (hasObject(address))
+        {
+            object_[address].size         = size;
+            object_[address].Ls           = Ls;
+            object_[address].isRegistered = true;
+        }
+        else
+        {
+            HADRON_ERROR("no object with address " + std::to_string(address));
+        }
+    }
+    else
+    {
+        HADRON_ERROR("object with address " + std::to_string(address)
+                     + " already registered");
+    }
+}
+
+void Environment::registerObject(const std::string name,
+                                 const unsigned int size, const unsigned int Ls)
+{
+    if (!hasObject(name))
+    {
+        addObject(name);
+    }
+    registerObject(getObjectAddress(name), size, Ls);
+}
+
+unsigned int Environment::getObjectAddress(const std::string name) const
+{
+    if (hasObject(name))
+    {
+        return objectAddress_.at(name);
+    }
+    else
+    {
+        HADRON_ERROR("no object with name '" + name + "'");
+    }
+}
+
+std::string Environment::getObjectName(const unsigned int address) const
+{
+    if (hasObject(address))
+    {
+        return object_[address].name;
+    }
+    else
+    {
+        HADRON_ERROR("no object with address " + std::to_string(address));
+    }
+}
+
+std::string Environment::getObjectType(const unsigned int address) const
+{
+    if (hasRegisteredObject(address))
+    {
+        if (object_[address].type)
+        {
+            return typeName(object_[address].type);
+        }
+        else
+        {
+            return "<no type>";
+        }
+    }
+    else if (hasObject(address))
+    {
+        HADRON_ERROR("object with address " + std::to_string(address)
+                     + " exists but is not registered");
+    }
+    else
+    {
+        HADRON_ERROR("no object with address " + std::to_string(address));
+    }
+}
+
+std::string Environment::getObjectType(const std::string name) const
+{
+    return getObjectType(getObjectAddress(name));
+}
+
+Environment::Size Environment::getObjectSize(const unsigned int address) const
+{
+    if (hasRegisteredObject(address))
+    {
+        return object_[address].size;
+    }
+    else if (hasObject(address))
+    {
+        HADRON_ERROR("object with address " + std::to_string(address)
+                     + " exists but is not registered");
+    }
+    else
+    {
+        HADRON_ERROR("no object with address " + std::to_string(address));
+    }
+}
+
+Environment::Size Environment::getObjectSize(const std::string name) const
+{
+    return getObjectSize(getObjectAddress(name));
+}
+
+unsigned int Environment::getObjectModule(const unsigned int address) const
+{
+    if (hasObject(address))
+    {
+        return object_[address].module;
+    }
+    else
+    {
+        HADRON_ERROR("no object with address " + std::to_string(address));
+    }
+}
+
+unsigned int Environment::getObjectModule(const std::string name) const
+{
+    return getObjectModule(getObjectAddress(name));
+}
+
+unsigned int Environment::getObjectLs(const unsigned int address) const
+{
+    if (hasRegisteredObject(address))
+    {
+        return object_[address].Ls;
+    }
+    else if (hasObject(address))
+    {
+        HADRON_ERROR("object with address " + std::to_string(address)
+                     + " exists but is not registered");
+    }
+    else
+    {
+        HADRON_ERROR("no object with address " + std::to_string(address));
+    }
+}
+
+unsigned int Environment::getObjectLs(const std::string name) const
+{
+    return getObjectLs(getObjectAddress(name));
+}
+
+bool Environment::hasObject(const unsigned int address) const
+{
+    return (address < object_.size());
+}
+
+bool Environment::hasObject(const std::string name) const
+{
+    auto it = objectAddress_.find(name);
+    
+    return ((it != objectAddress_.end()) and hasObject(it->second));
+}
+
+bool Environment::hasRegisteredObject(const unsigned int address) const
+{
+    if (hasObject(address))
+    {
+        return object_[address].isRegistered;
+    }
+    else
+    {
+        return false;
+    }
+}
+
+bool Environment::hasRegisteredObject(const std::string name) const
+{
+    if (hasObject(name))
+    {
+        return hasRegisteredObject(getObjectAddress(name));
+    }
+    else
+    {
+        return false;
+    }
+}
+
+bool Environment::hasCreatedObject(const unsigned int address) const
+{
+    if (hasObject(address))
+    {
+        return (object_[address].data != nullptr);
+    }
+    else
+    {
+        return false;
+    }
+}
+
+bool Environment::hasCreatedObject(const std::string name) const
+{
+    if (hasObject(name))
+    {
+        return hasCreatedObject(getObjectAddress(name));
+    }
+    else
+    {
+        return false;
+    }
+}
+
+bool Environment::isObject5d(const unsigned int address) const
+{
+    return (getObjectLs(address) > 1);
+}
+
+bool Environment::isObject5d(const std::string name) const
+{
+    return (getObjectLs(name) > 1);
+}
+
+Environment::Size Environment::getTotalSize(void) const
+{
+    Environment::Size size = 0;
+    
+    for (auto &o: object_)
+    {
+        if (o.isRegistered)
+        {
+            size += o.size;
+        }
+    }
+    
+    return size;
+}
+
+void Environment::addOwnership(const unsigned int owner,
+                               const unsigned int property)
+{
+    if (hasObject(property))
+    {
+        object_[property].owners.insert(owner);
+    }
+    else
+    {
+        HADRON_ERROR("no object with address " + std::to_string(property));
+    }
+    if (hasObject(owner))
+    {
+        object_[owner].properties.insert(property);
+    }
+    else
+    {
+        HADRON_ERROR("no object with address " + std::to_string(owner));
+    }
+}
+
+void Environment::addOwnership(const std::string owner,
+                               const std::string property)
+{
+    addOwnership(getObjectAddress(owner), getObjectAddress(property));
+}
+
+bool Environment::hasOwners(const unsigned int address) const
+{
+    
+    if (hasObject(address))
+    {
+        return (!object_[address].owners.empty());
+    }
+    else
+    {
+        HADRON_ERROR("no object with address " + std::to_string(address));
+    }
+}
+
+bool Environment::hasOwners(const std::string name) const
+{
+    return hasOwners(getObjectAddress(name));
+}
+
+bool Environment::freeObject(const unsigned int address)
+{
+    if (!hasOwners(address))
+    {
+        if (!isDryRun() and object_[address].isRegistered)
+        {
+            LOG(Message) << "Destroying object '" << object_[address].name
+                         << "'" << std::endl;
+        }
+        for (auto &p: object_[address].properties)
+        {
+            object_[p].owners.erase(address);
+        }
+        object_[address].size         = 0;
+        object_[address].Ls           = 0;
+        object_[address].isRegistered = false;
+        object_[address].type         = nullptr;
+        object_[address].owners.clear();
+        object_[address].properties.clear();
+        object_[address].data.reset(nullptr);
+        
+        return true;
+    }
+    else
+    {
+        return false;
+    }
+}
+
+bool Environment::freeObject(const std::string name)
+{
+    return freeObject(getObjectAddress(name));
+}
+
+void Environment::freeAll(void)
+{
+    for (unsigned int i = 0; i < object_.size(); ++i)
+    {
+        freeObject(i);
+    }
+}
+
+void Environment::printContent(void)
+{
+    LOG(Message) << "Modules: " << std::endl;
+    for (unsigned int i = 0; i < module_.size(); ++i)
+    {
+        LOG(Message) << std::setw(4) << i << ": "
+                     << getModuleName(i) << std::endl;
+    }
+    LOG(Message) << "Objects: " << std::endl;
+    for (unsigned int i = 0; i < object_.size(); ++i)
+    {
+        LOG(Message) << std::setw(4) << i << ": "
+                     << getObjectName(i) << std::endl;
+    }
+}

extras/Hadrons/Environment.hpp

@@ -0,0 +1,427 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Environment.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_Environment_hpp_
+#define Hadrons_Environment_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Graph.hpp>
+
+#ifndef SITE_SIZE_TYPE
+#define SITE_SIZE_TYPE unsigned int
+#endif
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         Global environment                                 *
+ ******************************************************************************/
+// forward declaration of Module
+class ModuleBase;
+
+class Object
+{
+public:
+    Object(void) = default;
+    virtual ~Object(void) = default;
+};
+
+template <typename T>
+class Holder: public Object
+{
+public:
+    Holder(void) = default;
+    Holder(T *pt);
+    virtual ~Holder(void) = default;
+    T &       get(void) const;
+    T *       getPt(void) const;
+    void      reset(T *pt);
+private:
+    std::unique_ptr<T> objPt_{nullptr};
+};
+
+class Environment
+{
+    SINGLETON(Environment);
+public:
+    typedef SITE_SIZE_TYPE                         Size;
+    typedef std::unique_ptr<ModuleBase>            ModPt;
+    typedef std::unique_ptr<GridCartesian>         GridPt;
+    typedef std::unique_ptr<GridRedBlackCartesian> GridRbPt;
+    typedef std::unique_ptr<GridParallelRNG>       RngPt;
+    typedef std::unique_ptr<LatticeBase>           LatticePt;
+private:
+    struct ModuleInfo
+    {
+        const std::type_info      *type{nullptr};
+        std::string               name;
+        ModPt                     data{nullptr};
+        std::vector<unsigned int> input;
+    };
+    struct ObjInfo
+    {
+        Size                    size{0};
+        unsigned int            Ls{0};
+        bool                    isRegistered{false};
+        const std::type_info    *type{nullptr};
+        std::string             name;
+        int                     module{-1};
+        std::set<unsigned int>  owners, properties;
+        std::unique_ptr<Object> data{nullptr};
+    };
+public:
+    // dry run
+    void                    dryRun(const bool isDry);
+    bool                    isDryRun(void) const;
+    // trajectory number
+    void                    setTrajectory(const unsigned int traj);
+    unsigned int            getTrajectory(void) const;
+    // grids
+    void                    createGrid(const unsigned int Ls);
+    GridCartesian *         getGrid(const unsigned int Ls = 1) const;
+    GridRedBlackCartesian * getRbGrid(const unsigned int Ls = 1) const;
+    std::vector<int>        getDim(void) const;
+    int                     getDim(const unsigned int mu) const;
+    unsigned int            getNd(void) const;
+    // random number generator
+    void                    setSeed(const std::vector<int> &seed);
+    GridParallelRNG *       get4dRng(void) const;
+    // module management
+    void                    pushModule(ModPt &pt);
+    template <typename M>
+    void                    createModule(const std::string name);
+    template <typename M>
+    void                    createModule(const std::string name,
+                                         const typename M::Par &par);
+    void                    createModule(const std::string name,
+                                         const std::string type,
+                                         XmlReader &reader);
+    unsigned int            getNModule(void) const;
+    ModuleBase *            getModule(const unsigned int address) const;
+    ModuleBase *            getModule(const std::string name) const;
+    template <typename M>
+    M *                     getModule(const unsigned int address) const;
+    template <typename M>
+    M *                     getModule(const std::string name) const;
+    unsigned int            getModuleAddress(const std::string name) const;
+    std::string             getModuleName(const unsigned int address) const;
+    std::string             getModuleType(const unsigned int address) const;
+    std::string             getModuleType(const std::string name) const;
+    std::string             getModuleNamespace(const unsigned int address) const;
+    std::string             getModuleNamespace(const std::string name) const;
+    bool                    hasModule(const unsigned int address) const;
+    bool                    hasModule(const std::string name) const;
+    Graph<unsigned int>     makeModuleGraph(void) const;
+    Size                    executeProgram(const std::vector<unsigned int> &p);
+    Size                    executeProgram(const std::vector<std::string> &p);
+    // general memory management
+    void                    addObject(const std::string name,
+                                      const int moduleAddress = -1);
+    void                    registerObject(const unsigned int address,
+                                           const unsigned int size,
+                                           const unsigned int Ls = 1);
+    void                    registerObject(const std::string name,
+                                           const unsigned int size,
+                                           const unsigned int Ls = 1);
+    template <typename T>
+    unsigned int            lattice4dSize(void) const;
+    template <typename T>
+    void                    registerLattice(const unsigned int address,
+                                            const unsigned int Ls = 1);
+    template <typename T>
+    void                    registerLattice(const std::string name,
+                                            const unsigned int Ls = 1);
+    template <typename T>
+    void                    setObject(const unsigned int address, T *object);
+    template <typename T>
+    void                    setObject(const std::string name, T *object);
+    template <typename T>
+    T *                     getObject(const unsigned int address) const;
+    template <typename T>
+    T *                     getObject(const std::string name) const;
+    template <typename T>
+    T *                     createLattice(const unsigned int address);
+    template <typename T>
+    T *                     createLattice(const std::string name);
+    unsigned int            getObjectAddress(const std::string name) const;
+    std::string             getObjectName(const unsigned int address) const;
+    std::string             getObjectType(const unsigned int address) const;
+    std::string             getObjectType(const std::string name) const;
+    Size                    getObjectSize(const unsigned int address) const;
+    Size                    getObjectSize(const std::string name) const;
+    unsigned int            getObjectModule(const unsigned int address) const;
+    unsigned int            getObjectModule(const std::string name) const;
+    unsigned int            getObjectLs(const unsigned int address) const;
+    unsigned int            getObjectLs(const std::string name) const;
+    bool                    hasObject(const unsigned int address) const;
+    bool                    hasObject(const std::string name) const;
+    bool                    hasRegisteredObject(const unsigned int address) const;
+    bool                    hasRegisteredObject(const std::string name) const;
+    bool                    hasCreatedObject(const unsigned int address) const;
+    bool                    hasCreatedObject(const std::string name) const;
+    bool                    isObject5d(const unsigned int address) const;
+    bool                    isObject5d(const std::string name) const;
+    template <typename T>
+    bool                    isObjectOfType(const unsigned int address) const;
+    template <typename T>
+    bool                    isObjectOfType(const std::string name) const;
+    Environment::Size       getTotalSize(void) const;
+    void                    addOwnership(const unsigned int owner,
+                                         const unsigned int property);
+    void                    addOwnership(const std::string owner,
+                                         const std::string property);
+    bool                    hasOwners(const unsigned int address) const;
+    bool                    hasOwners(const std::string name) const;
+    bool                    freeObject(const unsigned int address);
+    bool                    freeObject(const std::string name);
+    void                    freeAll(void);
+    void                    printContent(void);
+private:
+    // general
+    bool                                   dryRun_{false};
+    unsigned int                           traj_, locVol_;
+    // grids
+    std::vector<int>                       dim_;
+    GridPt                                 grid4d_;
+    std::map<unsigned int, GridPt>         grid5d_;
+    GridRbPt                               gridRb4d_;
+    std::map<unsigned int, GridRbPt>       gridRb5d_;
+    unsigned int                           nd_;
+    // random number generator
+    RngPt                                  rng4d_;
+    // module and related maps
+    std::vector<ModuleInfo>                module_;
+    std::map<std::string, unsigned int>    moduleAddress_;
+    // lattice store
+    std::map<unsigned int, LatticePt>      lattice_;
+    // object store
+    std::vector<ObjInfo>                   object_;
+    std::map<std::string, unsigned int>    objectAddress_;
+};
+
+/******************************************************************************
+ *                       Holder template implementation                       *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename T>
+Holder<T>::Holder(T *pt)
+: objPt_(pt)
+{}
+
+// access //////////////////////////////////////////////////////////////////////
+template <typename T>
+T & Holder<T>::get(void) const
+{
+    return &objPt_.get();
+}
+
+template <typename T>
+T * Holder<T>::getPt(void) const
+{
+    return objPt_.get();
+}
+
+template <typename T>
+void Holder<T>::reset(T *pt)
+{
+    objPt_.reset(pt);
+}
+
+/******************************************************************************
+ *                     Environment template implementation                    *
+ ******************************************************************************/
+// module management ///////////////////////////////////////////////////////////
+template <typename M>
+void Environment::createModule(const std::string name)
+{
+    ModPt pt(new M(name));
+    
+    pushModule(pt);
+}
+
+template <typename M>
+void Environment::createModule(const std::string name,
+                               const typename M::Par &par)
+{
+    ModPt pt(new M(name));
+    
+    static_cast<M *>(pt.get())->setPar(par);
+    pushModule(pt);
+}
+
+template <typename M>
+M * Environment::getModule(const unsigned int address) const
+{
+    if (auto *pt = dynamic_cast<M *>(getModule(address)))
+    {
+        return pt;
+    }
+    else
+    {
+        HADRON_ERROR("module '" + module_[address].name
+                     + "' does not have type " + typeid(M).name()
+                     + "(object type: " + getModuleType(address) + ")");
+    }
+}
+
+template <typename M>
+M * Environment::getModule(const std::string name) const
+{
+    return getModule<M>(getModuleAddress(name));
+}
+
+template <typename T>
+unsigned int Environment::lattice4dSize(void) const
+{
+    return sizeof(typename T::vector_object)/getGrid()->Nsimd();
+}
+
+template <typename T>
+void Environment::registerLattice(const unsigned int address,
+                                  const unsigned int Ls)
+{
+    createGrid(Ls);
+    registerObject(address, Ls*lattice4dSize<T>(), Ls);
+}
+
+template <typename T>
+void Environment::registerLattice(const std::string name, const unsigned int Ls)
+{
+    createGrid(Ls);
+    registerObject(name, Ls*lattice4dSize<T>(), Ls);
+}
+
+template <typename T>
+void Environment::setObject(const unsigned int address, T *object)
+{
+    if (hasRegisteredObject(address))
+    {
+        object_[address].data.reset(new Holder<T>(object));
+        object_[address].type = &typeid(T);
+    }
+    else if (hasObject(address))
+    {
+        HADRON_ERROR("object with address " + std::to_string(address) +
+                     " exists but is not registered");
+    }
+    else
+    {
+        HADRON_ERROR("no object with address " + std::to_string(address));
+    }
+}
+
+template <typename T>
+void Environment::setObject(const std::string name, T *object)
+{
+    setObject(getObjectAddress(name), object);
+}
+
+template <typename T>
+T * Environment::getObject(const unsigned int address) const
+{
+    if (hasRegisteredObject(address))
+    {
+        if (auto h = dynamic_cast<Holder<T> *>(object_[address].data.get()))
+        {
+            return h->getPt();
+        }
+        else
+        {
+            HADRON_ERROR("object with address " + std::to_string(address) +
+                         " does not have type '" + typeName(&typeid(T)) +
+                         "' (has type '" + getObjectType(address) + "')");
+        }
+    }
+    else if (hasObject(address))
+    {
+        HADRON_ERROR("object with address " + std::to_string(address) +
+                     " exists but is not registered");
+    }
+    else
+    {
+        HADRON_ERROR("no object with address " + std::to_string(address));
+    }
+}
+
+template <typename T>
+T * Environment::getObject(const std::string name) const
+{
+    return getObject<T>(getObjectAddress(name));
+}
+
+template <typename T>
+T * Environment::createLattice(const unsigned int address)
+{
+    GridCartesian *g = getGrid(getObjectLs(address));
+    
+    setObject(address, new T(g));
+    
+    return getObject<T>(address);
+}
+
+template <typename T>
+T * Environment::createLattice(const std::string name)
+{
+    return createLattice<T>(getObjectAddress(name));
+}
+
+template <typename T>
+bool Environment::isObjectOfType(const unsigned int address) const
+{
+    if (hasRegisteredObject(address))
+    {
+        if (auto h = dynamic_cast<Holder<T> *>(object_[address].data.get()))
+        {
+            return true;
+        }
+        else
+        {
+            return false;
+        }
+    }
+    else if (hasObject(address))
+    {
+        HADRON_ERROR("object with address " + std::to_string(address) +
+                     " exists but is not registered");
+    }
+    else
+    {
+        HADRON_ERROR("no object with address " + std::to_string(address));
+    }
+}
+
+template <typename T>
+bool Environment::isObjectOfType(const std::string name) const
+{
+    return isObjectOfType<T>(getObjectAddress(name));
+}
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_Environment_hpp_

extras/Hadrons/Factory.hpp

@@ -0,0 +1,106 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Factory.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_Factory_hpp_
+#define Hadrons_Factory_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                        abstract factory class                              *
+ ******************************************************************************/
+template <typename T>
+class Factory
+{
+public:
+    typedef std::function<std::unique_ptr<T>(const std::string)> Func;
+public:
+    // constructor
+    Factory(void) = default;
+    // destructor
+    virtual ~Factory(void) = default;
+    // registration
+    void registerBuilder(const std::string type, const Func &f);
+    // get builder list
+    std::vector<std::string> getBuilderList(void) const;
+    // factory
+    std::unique_ptr<T> create(const std::string type,
+                              const std::string name) const;
+private:
+    std::map<std::string, Func> builder_;
+};
+
+/******************************************************************************
+ *                         template implementation                            *
+ ******************************************************************************/
+// registration ////////////////////////////////////////////////////////////////
+template <typename T>
+void Factory<T>::registerBuilder(const std::string type, const Func &f)
+{
+    builder_[type] = f;
+}
+
+// get module list /////////////////////////////////////////////////////////////
+template <typename T>
+std::vector<std::string> Factory<T>::getBuilderList(void) const
+{
+    std::vector<std::string> list;
+    
+    for (auto &b: builder_)
+    {
+        list.push_back(b.first);
+    }
+    
+    return list;
+}
+
+// factory /////////////////////////////////////////////////////////////////////
+template <typename T>
+std::unique_ptr<T> Factory<T>::create(const std::string type,
+                                      const std::string name) const
+{
+    Func func;
+    
+    try
+    {
+        func = builder_.at(type);
+    }
+    catch (std::out_of_range &)
+    {
+        HADRON_ERROR("object of type '" + type + "' unknown");
+    }
+    
+    return func(name);
+}
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_Factory_hpp_

extras/Hadrons/GeneticScheduler.hpp

@@ -0,0 +1,329 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/GeneticScheduler.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_GeneticScheduler_hpp_
+#define Hadrons_GeneticScheduler_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Graph.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                   Scheduler based on a genetic algorithm                   *
+ ******************************************************************************/
+template <typename T>
+class GeneticScheduler
+{
+public:
+    typedef std::vector<T>                   Gene;
+    typedef std::pair<Gene *, Gene *>        GenePair;
+    typedef std::function<int(const Gene &)> ObjFunc;
+    struct Parameters
+    {
+        double       mutationRate;
+        unsigned int popSize, seed;
+    };
+public:
+    // constructor
+    GeneticScheduler(Graph<T> &graph, const ObjFunc &func,
+                     const Parameters &par);
+    // destructor
+    virtual ~GeneticScheduler(void) = default;
+    // access
+    const Gene & getMinSchedule(void);
+    int          getMinValue(void);
+    // breed a new generation
+    void nextGeneration(void);
+    // heuristic benchmarks
+    void benchmarkCrossover(const unsigned int nIt);
+    // print population
+    friend std::ostream & operator<<(std::ostream &out,
+                                     const GeneticScheduler<T> &s)
+    {
+        out << "[";
+        for (auto &p: s.population_)
+        {
+            out << p.first << ", ";
+        }
+        out << "\b\b]";
+        
+        return out;
+    }
+private:
+    // evolution steps
+    void initPopulation(void);
+    void doCrossover(void);
+    void doMutation(void);
+    // genetic operators
+    GenePair selectPair(void);
+    void     crossover(Gene &c1, Gene &c2, const Gene &p1, const Gene &p2);
+    void     mutation(Gene &m, const Gene &c);
+    
+private:
+    Graph<T>                 &graph_;
+    const ObjFunc            &func_;
+    const Parameters         par_;
+    std::multimap<int, Gene> population_;
+    std::mt19937             gen_;
+};
+
+/******************************************************************************
+ *                       template implementation                              *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename T>
+GeneticScheduler<T>::GeneticScheduler(Graph<T> &graph, const ObjFunc &func,
+                                      const Parameters &par)
+: graph_(graph)
+, func_(func)
+, par_(par)
+{
+    gen_.seed(par_.seed);
+}
+
+// access //////////////////////////////////////////////////////////////////////
+template <typename T>
+const typename GeneticScheduler<T>::Gene &
+GeneticScheduler<T>::getMinSchedule(void)
+{
+    return population_.begin()->second;
+}
+
+template <typename T>
+int GeneticScheduler<T>::getMinValue(void)
+{
+    return population_.begin()->first;
+}
+
+// breed a new generation //////////////////////////////////////////////////////
+template <typename T>
+void GeneticScheduler<T>::nextGeneration(void)
+{
+    // random initialization of the population if necessary
+    if (population_.size() != par_.popSize)
+    {
+        initPopulation();
+    }
+    LOG(Debug) << "Starting population:\n" << *this << std::endl;
+    
+    // random mutations
+    //PARALLEL_FOR_LOOP
+    for (unsigned int i = 0; i < par_.popSize; ++i)
+    {
+        doMutation();
+    }
+    LOG(Debug) << "After mutations:\n" << *this << std::endl;
+    
+    // mating
+    //PARALLEL_FOR_LOOP
+    for (unsigned int i = 0; i < par_.popSize/2; ++i)
+    {
+        doCrossover();
+    }
+    LOG(Debug) << "After mating:\n" << *this << std::endl;
+    
+    // grim reaper
+    auto it = population_.begin();
+    
+    std::advance(it, par_.popSize);
+    population_.erase(it, population_.end());
+    LOG(Debug) << "After grim reaper:\n" << *this << std::endl;
+}
+
+// evolution steps /////////////////////////////////////////////////////////////
+template <typename T>
+void GeneticScheduler<T>::initPopulation(void)
+{
+    population_.clear();
+    for (unsigned int i = 0; i < par_.popSize; ++i)
+    {
+        auto p = graph_.topoSort(gen_);
+        
+        population_.insert(std::make_pair(func_(p), p));
+    }
+}
+
+template <typename T>
+void GeneticScheduler<T>::doCrossover(void)
+{
+    auto p = selectPair();
+    Gene &p1 = *(p.first), &p2 = *(p.second);
+    Gene c1, c2;
+    
+    crossover(c1, c2, p1, p2);
+    PARALLEL_CRITICAL
+    {
+        population_.insert(std::make_pair(func_(c1), c1));
+        population_.insert(std::make_pair(func_(c2), c2));
+    }
+}
+
+template <typename T>
+void GeneticScheduler<T>::doMutation(void)
+{
+    std::uniform_real_distribution<double>      mdis(0., 1.);
+    std::uniform_int_distribution<unsigned int> pdis(0, population_.size() - 1);
+    
+    if (mdis(gen_) < par_.mutationRate)
+    {
+        Gene m;
+        auto it = population_.begin();
+        
+        std::advance(it, pdis(gen_));
+        mutation(m, it->second);
+        PARALLEL_CRITICAL
+        {
+            population_.insert(std::make_pair(func_(m), m));
+        }
+    }
+}
+
+// genetic operators ///////////////////////////////////////////////////////////
+template <typename T>
+typename GeneticScheduler<T>::GenePair GeneticScheduler<T>::selectPair(void)
+{
+    std::vector<double> prob;
+    unsigned int        ind;
+    Gene                *p1, *p2;
+    
+    for (auto &c: population_)
+    {
+        prob.push_back(1./c.first);
+    }
+    do
+    {
+        double probCpy;
+        
+        std::discrete_distribution<unsigned int> dis1(prob.begin(), prob.end());
+        auto rIt = population_.begin();
+        ind = dis1(gen_);
+        std::advance(rIt, ind);
+        p1 = &(rIt->second);
+        probCpy   = prob[ind];
+        prob[ind] = 0.;
+        std::discrete_distribution<unsigned int> dis2(prob.begin(), prob.end());
+        rIt = population_.begin();
+        std::advance(rIt, dis2(gen_));
+        p2 = &(rIt->second);
+        prob[ind] = probCpy;
+    } while (p1 == p2);
+    
+    return std::make_pair(p1, p2);
+}
+
+template <typename T>
+void GeneticScheduler<T>::crossover(Gene &c1, Gene &c2, const Gene &p1,
+                                    const Gene &p2)
+{
+    Gene                                        buf;
+    std::uniform_int_distribution<unsigned int> dis(0, p1.size() - 1);
+    unsigned int                                cut = dis(gen_);
+    
+    c1.clear();
+    buf = p2;
+    for (unsigned int i = 0; i < cut; ++i)
+    {
+        c1.push_back(p1[i]);
+        buf.erase(std::find(buf.begin(), buf.end(), p1[i]));
+    }
+    for (unsigned int i = 0; i < buf.size(); ++i)
+    {
+        c1.push_back(buf[i]);
+    }
+    c2.clear();
+    buf = p2;
+    for (unsigned int i = cut; i < p1.size(); ++i)
+    {
+        buf.erase(std::find(buf.begin(), buf.end(), p1[i]));
+    }
+    for (unsigned int i = 0; i < buf.size(); ++i)
+    {
+        c2.push_back(buf[i]);
+    }
+    for (unsigned int i = cut; i < p1.size(); ++i)
+    {
+        c2.push_back(p1[i]);
+    }
+}
+
+template <typename T>
+void GeneticScheduler<T>::mutation(Gene &m, const Gene &c)
+{
+    Gene                                        buf;
+    std::uniform_int_distribution<unsigned int> dis(0, c.size() - 1);
+    unsigned int                                cut = dis(gen_);
+    Graph<T>                                    g1 = graph_, g2 = graph_;
+    
+    for (unsigned int i = 0; i < cut; ++i)
+    {
+        g1.removeVertex(c[i]);
+    }
+    for (unsigned int i = cut; i < c.size(); ++i)
+    {
+        g2.removeVertex(c[i]);
+    }
+    if (g1.size() > 0)
+    {
+        buf = g1.topoSort(gen_);
+    }
+    if (g2.size() > 0)
+    {
+        m = g2.topoSort(gen_);
+    }
+    for (unsigned int i = cut; i < c.size(); ++i)
+    {
+        m.push_back(buf[i - cut]);
+    }
+}
+
+template <typename T>
+void GeneticScheduler<T>::benchmarkCrossover(const unsigned int nIt)
+{
+    Gene   p1, p2, c1, c2;
+    double neg = 0., eq = 0., pos = 0., total;
+    int    improvement;
+    
+    LOG(Message) << "Benchmarking crossover..." << std::endl;
+    for (unsigned int i = 0; i < nIt; ++i)
+    {
+        p1 = graph_.topoSort(gen_);
+        p2 = graph_.topoSort(gen_);
+        crossover(c1, c2, p1, p2);
+        improvement = (func_(c1) + func_(c2) - func_(p1) - func_(p2))/2;
+        if (improvement < 0) neg++; else if (improvement == 0) eq++; else pos++;
+    }
+    total = neg + eq + pos;
+    LOG(Message) << "  -: " << neg/total << "  =: " << eq/total
+                 << "  +: " << pos/total << std::endl;
+}
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_GeneticScheduler_hpp_

extras/Hadrons/Global.cc

@@ -0,0 +1,82 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Global.cc
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Hadrons/Global.hpp>
+
+using namespace Grid;
+using namespace QCD;
+using namespace Hadrons;
+
+HadronsLogger Hadrons::HadronsLogError(1,"Error");
+HadronsLogger Hadrons::HadronsLogWarning(1,"Warning");
+HadronsLogger Hadrons::HadronsLogMessage(1,"Message");
+HadronsLogger Hadrons::HadronsLogIterative(1,"Iterative");
+HadronsLogger Hadrons::HadronsLogDebug(1,"Debug");
+
+// pretty size formatting //////////////////////////////////////////////////////
+std::string Hadrons::sizeString(long unsigned int bytes)
+
+{
+    constexpr unsigned int bufSize = 256;
+    const char             *suffixes[7] = {"", "K", "M", "G", "T", "P", "E"};
+    char                   buf[256];
+    long unsigned int      s     = 0;
+    double                 count = bytes;
+    
+    while (count >= 1024 && s < 7)
+    {
+        s++;
+        count /= 1024;
+    }
+    if (count - floor(count) == 0.0)
+    {
+        snprintf(buf, bufSize, "%d %sB", (int)count, suffixes[s]);
+    }
+    else
+    {
+        snprintf(buf, bufSize, "%.1f %sB", count, suffixes[s]);
+    }
+    
+    return std::string(buf);
+}
+
+// type utilities //////////////////////////////////////////////////////////////
+constexpr unsigned int maxNameSize = 1024u;
+
+std::string Hadrons::typeName(const std::type_info *info)
+{
+    char        *buf;
+    std::string name;
+    
+    buf  = abi::__cxa_demangle(info->name(), nullptr, nullptr, nullptr);
+    name = buf;
+    free(buf);
+    
+    return name;
+}

extras/Hadrons/Global.hpp

@@ -0,0 +1,179 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Global.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_Global_hpp_
+#define Hadrons_Global_hpp_
+
+#include <set>
+#include <stack>
+#include <Grid/Grid.h>
+#include <cxxabi.h>
+
+#define BEGIN_HADRONS_NAMESPACE \
+namespace Grid {\
+using namespace QCD;\
+namespace Hadrons {\
+using Grid::operator<<;
+#define END_HADRONS_NAMESPACE }}
+
+#define BEGIN_MODULE_NAMESPACE(name)\
+namespace name {\
+using Grid::operator<<;
+#define END_MODULE_NAMESPACE }
+
+/* the 'using Grid::operator<<;' statement prevents a very nasty compilation
+ * error with GCC 5 (clang & GCC 6 compile fine without it).
+ */
+
+#ifndef FIMPL
+#define FIMPL WilsonImplR
+#endif
+#ifndef SIMPL
+#define SIMPL ScalarImplCR
+#endif
+
+BEGIN_HADRONS_NAMESPACE
+
+// type aliases
+#define FERM_TYPE_ALIASES(FImpl, suffix)\
+typedef FermionOperator<FImpl>                       FMat##suffix;             \
+typedef typename FImpl::FermionField                 FermionField##suffix;     \
+typedef typename FImpl::PropagatorField              PropagatorField##suffix;  \
+typedef typename FImpl::SitePropagator               SitePropagator##suffix;   \
+typedef std::vector<typename FImpl::SitePropagator::scalar_object>             \
+                                                     SlicedPropagator##suffix;
+
+#define GAUGE_TYPE_ALIASES(FImpl, suffix)\
+typedef typename FImpl::DoubledGaugeField DoubledGaugeField##suffix;
+
+#define SCALAR_TYPE_ALIASES(SImpl, suffix)\
+typedef typename SImpl::Field ScalarField##suffix;\
+typedef typename SImpl::Field PropagatorField##suffix;
+
+#define SOLVER_TYPE_ALIASES(FImpl, suffix)\
+typedef std::function<void(FermionField##suffix &,\
+                      const FermionField##suffix &)> SolverFn##suffix;
+
+#define SINK_TYPE_ALIASES(suffix)\
+typedef std::function<SlicedPropagator##suffix(const PropagatorField##suffix &)> SinkFn##suffix;
+
+#define FGS_TYPE_ALIASES(FImpl, suffix)\
+FERM_TYPE_ALIASES(FImpl, suffix)\
+GAUGE_TYPE_ALIASES(FImpl, suffix)\
+SOLVER_TYPE_ALIASES(FImpl, suffix)
+
+// logger
+class HadronsLogger: public Logger
+{
+public:
+    HadronsLogger(int on, std::string nm): Logger("Hadrons", on, nm,
+                                                  GridLogColours, "BLACK"){};
+};
+
+#define LOG(channel) std::cout << HadronsLog##channel
+#define HADRON_ERROR(msg)\
+LOG(Error) << msg << " (" << __FUNCTION__ << " at " << __FILE__ << ":"\
+           << __LINE__ << ")" << std::endl;\
+abort();
+
+#define DEBUG_VAR(var) LOG(Debug) << #var << "= " << (var) << std::endl;
+
+extern HadronsLogger HadronsLogError;
+extern HadronsLogger HadronsLogWarning;
+extern HadronsLogger HadronsLogMessage;
+extern HadronsLogger HadronsLogIterative;
+extern HadronsLogger HadronsLogDebug;
+
+// singleton pattern
+#define SINGLETON(name)\
+public:\
+    name(const name &e) = delete;\
+    void operator=(const name &e) = delete;\
+    static name & getInstance(void)\
+    {\
+        static name e;\
+        return e;\
+    }\
+private:\
+    name(void);
+
+#define SINGLETON_DEFCTOR(name)\
+public:\
+    name(const name &e) = delete;\
+    void operator=(const name &e) = delete;\
+    static name & getInstance(void)\
+    {\
+        static name e;\
+        return e;\
+    }\
+private:\
+    name(void) = default;
+
+// pretty size formating
+std::string sizeString(long unsigned int bytes);
+
+// type utilities
+template <typename T>
+const std::type_info * typeIdPt(const T &x)
+{
+    return &typeid(x);
+}
+
+std::string typeName(const std::type_info *info);
+
+template <typename T>
+const std::type_info * typeIdPt(void)
+{
+    return &typeid(T);
+}
+
+template <typename T>
+std::string typeName(const T &x)
+{
+    return typeName(typeIdPt(x));
+}
+
+template <typename T>
+std::string typeName(void)
+{
+    return typeName(typeIdPt<T>());
+}
+
+// default writers/readers
+#ifdef HAVE_HDF5
+typedef Hdf5Reader CorrReader;
+typedef Hdf5Writer CorrWriter;
+#else
+typedef XmlReader CorrReader;
+typedef XmlWriter CorrWriter;
+#endif
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_Global_hpp_

extras/Hadrons/Graph.hpp

@@ -0,0 +1,760 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Graph.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_Graph_hpp_
+#define Hadrons_Graph_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                          Oriented graph class                              *
+ ******************************************************************************/
+// I/O for edges
+template <typename T>
+std::ostream & operator<<(std::ostream &out, const std::pair<T, T> &e)
+{
+    out << "\""  << e.first << "\" -> \"" << e.second << "\"";
+    
+    return out;
+}
+
+// main class
+template <typename T>
+class Graph
+{
+public:
+    typedef std::pair<T, T> Edge;
+public:
+    // constructor
+    Graph(void);
+    // destructor
+    virtual ~Graph(void) = default;
+    // access
+    void           addVertex(const T &value);
+    void           addEdge(const Edge &e);
+    void           addEdge(const T &start, const T &end);
+    std::vector<T> getVertices(void) const;
+    void           removeVertex(const T &value);
+    void           removeEdge(const Edge &e);
+    void           removeEdge(const T &start, const T &end);
+    unsigned int   size(void) const;
+    // tests
+    bool gotValue(const T &value) const;
+    // graph topological manipulations
+    std::vector<T>              getAdjacentVertices(const T &value) const;
+    std::vector<T>              getChildren(const T &value) const;
+    std::vector<T>              getParents(const T &value) const;
+    std::vector<T>              getRoots(void) const;
+    std::vector<Graph<T>>       getConnectedComponents(void) const;
+    std::vector<T>              topoSort(void);
+    template <typename Gen>
+    std::vector<T>              topoSort(Gen &gen);
+    std::vector<std::vector<T>> allTopoSort(void);
+    // I/O
+    friend std::ostream & operator<<(std::ostream &out, const Graph<T> &g)
+    {
+        out << "{";
+        for (auto &e: g.edgeSet_)
+        {
+            out << e << ", ";
+        }
+        if (g.edgeSet_.size() != 0)
+        {
+            out << "\b\b";
+        }
+        out << "}";
+        
+        return out;
+    }
+private:
+    // vertex marking
+    void      mark(const T &value, const bool doMark = true);
+    void      markAll(const bool doMark = true);
+    void      unmark(const T &value);
+    void      unmarkAll(void);
+    bool      isMarked(const T &value) const;
+    const T * getFirstMarked(const bool isMarked = true) const;
+    template <typename Gen>
+    const T * getRandomMarked(const bool isMarked, Gen &gen);
+    const T * getFirstUnmarked(void) const;
+    template <typename Gen>
+    const T * getRandomUnmarked(Gen &gen);
+    // prune marked/unmarked vertices
+    void removeMarked(const bool isMarked = true);
+    void removeUnmarked(void);
+    // depth-first search marking
+    void depthFirstSearch(void);
+    void depthFirstSearch(const T &root);
+private:
+    std::map<T, bool>  isMarked_;
+    std::set<Edge>     edgeSet_;
+};
+
+// build depedency matrix from topological sorts
+template <typename T>
+std::map<T, std::map<T, bool>>
+makeDependencyMatrix(const std::vector<std::vector<T>> &topSort);
+
+/******************************************************************************
+ *                       template implementation                              *
+ ******************************************************************************
+ * in all the following V is the number of vertex and E is the number of edge
+ * in the worst case E = V^2
+ */
+
+// constructor /////////////////////////////////////////////////////////////////
+template <typename T>
+Graph<T>::Graph(void)
+{}
+
+// access //////////////////////////////////////////////////////////////////////
+// complexity: log(V)
+template <typename T>
+void Graph<T>::addVertex(const T &value)
+{
+    isMarked_[value] = false;
+}
+
+// complexity: O(log(V))
+template <typename T>
+void Graph<T>::addEdge(const Edge &e)
+{
+    addVertex(e.first);
+    addVertex(e.second);
+    edgeSet_.insert(e);
+}
+
+// complexity: O(log(V))
+template <typename T>
+void Graph<T>::addEdge(const T &start, const T &end)
+{
+    addEdge(Edge(start, end));
+}
+
+template <typename T>
+std::vector<T> Graph<T>::getVertices(void) const
+{
+    std::vector<T> vertex;
+    
+    for (auto &v: isMarked_)
+    {
+        vertex.push_back(v.first);
+    }
+    
+    return vertex;
+}
+
+// complexity: O(V*log(V))
+template <typename T>
+void Graph<T>::removeVertex(const T &value)
+{
+    // remove vertex from the mark table
+    auto vIt = isMarked_.find(value);
+    
+    if (vIt != isMarked_.end())
+    {
+        isMarked_.erase(vIt);
+    }
+    else
+    {
+        HADRON_ERROR("vertex " << value << " does not exists");
+    }
+
+    // remove all edges containing the vertex
+    auto pred = [&value](const Edge &e)
+    {
+        return ((e.first == value) or (e.second == value));
+    };
+    auto eIt = find_if(edgeSet_.begin(), edgeSet_.end(), pred);
+    
+    while (eIt != edgeSet_.end())
+    {
+        edgeSet_.erase(eIt);
+        eIt = find_if(edgeSet_.begin(), edgeSet_.end(), pred);
+    }
+}
+
+// complexity: O(log(V))
+template <typename T>
+void Graph<T>::removeEdge(const Edge &e)
+{
+    auto eIt = edgeSet_.find(e);
+    
+    if (eIt != edgeSet_.end())
+    {
+        edgeSet_.erase(eIt);
+    }
+    else
+    {
+        HADRON_ERROR("edge "  << e << " does not exists");
+    }
+}
+
+// complexity: O(log(V))
+template <typename T>
+void Graph<T>::removeEdge(const T &start, const T &end)
+{
+    removeEdge(Edge(start, end));
+}
+
+// complexity: O(1)
+template <typename T>
+unsigned int Graph<T>::size(void) const
+{
+    return isMarked_.size();
+}
+
+// tests ///////////////////////////////////////////////////////////////////////
+// complexity: O(log(V))
+template <typename T>
+bool Graph<T>::gotValue(const T &value) const
+{
+    auto it = isMarked_.find(value);
+    
+    if (it == isMarked_.end())
+    {
+        return false;
+    }
+    else
+    {
+        return true;
+    }
+}
+
+// vertex marking //////////////////////////////////////////////////////////////
+// complexity: O(log(V))
+template <typename T>
+void Graph<T>::mark(const T &value, const bool doMark)
+{
+    if (gotValue(value))
+    {
+        isMarked_[value] = doMark;
+    }
+    else
+    {
+        HADRON_ERROR("vertex " << value << " does not exists");
+    }
+}
+
+// complexity: O(V*log(V))
+template <typename T>
+void Graph<T>::markAll(const bool doMark)
+{
+    for (auto &v: isMarked_)
+    {
+        mark(v.first, doMark);
+    }
+}
+
+// complexity: O(log(V))
+template <typename T>
+void Graph<T>::unmark(const T &value)
+{
+    mark(value, false);
+}
+
+// complexity: O(V*log(V))
+template <typename T>
+void Graph<T>::unmarkAll(void)
+{
+    markAll(false);
+}
+
+// complexity: O(log(V))
+template <typename T>
+bool Graph<T>::isMarked(const T &value) const
+{
+    if (gotValue(value))
+    {
+        return isMarked_.at(value);
+    }
+    else
+    {
+        HADRON_ERROR("vertex " << value << " does not exists");
+        
+        return false;
+    }
+}
+
+// complexity: O(log(V))
+template <typename T>
+const T * Graph<T>::getFirstMarked(const bool isMarked) const
+{
+    auto pred = [&isMarked](const std::pair<T, bool> &v)
+    {
+        return (v.second == isMarked);
+    };
+    auto vIt = std::find_if(isMarked_.begin(), isMarked_.end(), pred);
+    
+    if (vIt != isMarked_.end())
+    {
+        return &(vIt->first);
+    }
+    else
+    {
+        return nullptr;
+    }
+}
+
+// complexity: O(log(V))
+template <typename T>
+template <typename Gen>
+const T * Graph<T>::getRandomMarked(const bool isMarked, Gen &gen)
+{
+    auto pred = [&isMarked](const std::pair<T, bool> &v)
+    {
+        return (v.second == isMarked);
+    };
+    std::uniform_int_distribution<unsigned int> dis(0, size() - 1);
+    auto                                        rIt = isMarked_.begin();
+    
+    std::advance(rIt, dis(gen));
+    auto vIt = std::find_if(rIt, isMarked_.end(), pred);
+    if (vIt != isMarked_.end())
+    {
+        return &(vIt->first);
+    }
+    else
+    {
+        vIt = std::find_if(isMarked_.begin(), rIt, pred);
+        if (vIt != rIt)
+        {
+            return &(vIt->first);
+        }
+        else
+        {
+            return nullptr;
+        }
+    }
+}
+
+// complexity: O(log(V))
+template <typename T>
+const T * Graph<T>::getFirstUnmarked(void) const
+{
+    return getFirstMarked(false);
+}
+
+// complexity: O(log(V))
+template <typename T>
+template <typename Gen>
+const T * Graph<T>::getRandomUnmarked(Gen &gen)
+{
+    return getRandomMarked(false, gen);
+}
+
+// prune marked/unmarked vertices //////////////////////////////////////////////
+// complexity: O(V^2*log(V))
+template <typename T>
+void Graph<T>::removeMarked(const bool isMarked)
+{
+    auto isMarkedCopy = isMarked_;
+    
+    for (auto &v: isMarkedCopy)
+    {
+        if (v.second == isMarked)
+        {
+            removeVertex(v.first);
+        }
+    }
+}
+
+// complexity: O(V^2*log(V))
+template <typename T>
+void Graph<T>::removeUnmarked(void)
+{
+    removeMarked(false);
+}
+
+// depth-first search marking //////////////////////////////////////////////////
+// complexity: O(V*log(V))
+template <typename T>
+void Graph<T>::depthFirstSearch(void)
+{
+    depthFirstSearch(isMarked_.begin()->first);
+}
+
+// complexity: O(V*log(V))
+template <typename T>
+void Graph<T>::depthFirstSearch(const T &root)
+{
+    std::vector<T> adjacentVertex;
+    
+    mark(root);
+    adjacentVertex = getAdjacentVertices(root);
+    for (auto &v: adjacentVertex)
+    {
+        if (!isMarked(v))
+        {
+            depthFirstSearch(v);
+        }
+    }
+}
+
+// graph topological manipulations /////////////////////////////////////////////
+// complexity: O(V*log(V))
+template <typename T>
+std::vector<T> Graph<T>::getAdjacentVertices(const T &value) const
+{
+    std::vector<T> adjacentVertex;
+    
+    auto pred = [&value](const Edge &e)
+    {
+        return ((e.first == value) or (e.second == value));
+    };
+    auto eIt = find_if(edgeSet_.begin(), edgeSet_.end(), pred);
+    
+    while (eIt != edgeSet_.end())
+    {
+        if (eIt->first == value)
+        {
+            adjacentVertex.push_back((*eIt).second);
+        }
+        else if (eIt->second == value)
+        {
+            adjacentVertex.push_back((*eIt).first);
+        }
+        eIt = find_if(++eIt, edgeSet_.end(), pred);
+    }
+    
+    return adjacentVertex;
+}
+
+// complexity: O(V*log(V))
+template <typename T>
+std::vector<T> Graph<T>::getChildren(const T &value) const
+{
+    std::vector<T> child;
+    
+    auto pred = [&value](const Edge &e)
+    {
+        return (e.first == value);
+    };
+    auto eIt = find_if(edgeSet_.begin(), edgeSet_.end(), pred);
+    
+    while (eIt != edgeSet_.end())
+    {
+        child.push_back((*eIt).second);
+        eIt = find_if(++eIt, edgeSet_.end(), pred);
+    }
+    
+    return child;
+}
+
+// complexity: O(V*log(V))
+template <typename T>
+std::vector<T> Graph<T>::getParents(const T &value) const
+{
+    std::vector<T> parent;
+    
+    auto pred = [&value](const Edge &e)
+    {
+        return (e.second == value);
+    };
+    auto eIt = find_if(edgeSet_.begin(), edgeSet_.end(), pred);
+    
+    while (eIt != edgeSet_.end())
+    {
+        parent.push_back((*eIt).first);
+        eIt = find_if(++eIt, edgeSet_.end(), pred);
+    }
+    
+    return parent;
+}
+
+// complexity: O(V^2*log(V))
+template <typename T>
+std::vector<T> Graph<T>::getRoots(void) const
+{
+    std::vector<T> root;
+    
+    for (auto &v: isMarked_)
+    {
+        auto parent = getParents(v.first);
+        
+        if (parent.size() == 0)
+        {
+            root.push_back(v.first);
+        }
+    }
+    
+    return root;
+}
+
+// complexity: O(V^2*log(V))
+template <typename T>
+std::vector<Graph<T>> Graph<T>::getConnectedComponents(void) const
+{
+    std::vector<Graph<T>> res;
+    Graph<T>              copy(*this);
+    
+    while (copy.size() > 0)
+    {
+        copy.depthFirstSearch();
+        res.push_back(copy);
+        res.back().removeUnmarked();
+        res.back().unmarkAll();
+        copy.removeMarked();
+        copy.unmarkAll();
+    }
+    
+    return res;
+}
+
+// topological sort using a directed DFS algorithm
+// complexity: O(V*log(V))
+template <typename T>
+std::vector<T> Graph<T>::topoSort(void)
+{
+    std::stack<T>     buf;
+    std::vector<T>    res;
+    const T           *vPt;
+    std::map<T, bool> tmpMarked(isMarked_);
+
+    // visit function
+    std::function<void(const T &)> visit = [&](const T &v)
+    {
+        if (tmpMarked.at(v))
+        {
+            HADRON_ERROR("cannot topologically sort a cyclic graph");
+        }
+        if (!isMarked(v))
+        {
+            std::vector<T> child = getChildren(v);
+
+            tmpMarked[v] = true;
+            for (auto &c: child)
+            {
+                visit(c);
+            }
+            mark(v);
+            tmpMarked[v] = false;
+            buf.push(v);
+        }
+    };
+    
+    // reset temporary marks
+    for (auto &v: tmpMarked)
+    {
+        tmpMarked.at(v.first) = false;
+    }
+    
+    // loop on unmarked vertices
+    unmarkAll();
+    vPt = getFirstUnmarked();
+    while (vPt)
+    {
+        visit(*vPt);
+        vPt = getFirstUnmarked();
+    }
+    unmarkAll();
+    
+    // create result vector
+    while (!buf.empty())
+    {
+        res.push_back(buf.top());
+        buf.pop();
+    }
+    
+    return res;
+}
+
+// random version of the topological sort
+// complexity: O(V*log(V))
+template <typename T>
+template <typename Gen>
+std::vector<T> Graph<T>::topoSort(Gen &gen)
+{
+    std::stack<T>     buf;
+    std::vector<T>    res;
+    const T           *vPt;
+    std::map<T, bool> tmpMarked(isMarked_);
+    
+    // visit function
+    std::function<void(const T &)> visit = [&](const T &v)
+    {
+        if (tmpMarked.at(v))
+        {
+            HADRON_ERROR("cannot topologically sort a cyclic graph");
+        }
+        if (!isMarked(v))
+        {
+            std::vector<T> child = getChildren(v);
+            
+            tmpMarked[v] = true;
+            std::shuffle(child.begin(), child.end(), gen);
+            for (auto &c: child)
+            {
+                visit(c);
+            }
+            mark(v);
+            tmpMarked[v] = false;
+            buf.push(v);
+        }
+    };
+    
+    // reset temporary marks
+    for (auto &v: tmpMarked)
+    {
+        tmpMarked.at(v.first) = false;
+    }
+    
+    // loop on unmarked vertices
+    unmarkAll();
+    vPt = getRandomUnmarked(gen);
+    while (vPt)
+    {
+        visit(*vPt);
+        vPt = getRandomUnmarked(gen);
+    }
+    unmarkAll();
+    
+    // create result vector
+    while (!buf.empty())
+    {
+        res.push_back(buf.top());
+        buf.pop();
+    }
+    
+    return res;
+}
+
+// generate all possible topological sorts
+// Y. L. Varol & D. Rotem, Comput. J. 24(1), pp. 83–84, 1981
+// http://comjnl.oupjournals.org/cgi/doi/10.1093/comjnl/24.1.83
+// complexity: O(V*log(V)) (from the paper, but really ?)
+template <typename T>
+std::vector<std::vector<T>> Graph<T>::allTopoSort(void)
+{
+    std::vector<std::vector<T>>    res;
+    std::map<T, std::map<T, bool>> iMat;
+    
+    // create incidence matrix
+    for (auto &v1: isMarked_)
+    for (auto &v2: isMarked_)
+    {
+        iMat[v1.first][v2.first] = false;
+    }
+    for (auto &v: isMarked_)
+    {
+        auto cVec = getChildren(v.first);
+        
+        for (auto &c: cVec)
+        {
+            iMat[v.first][c] = true;
+        }
+    }
+    
+    // generate initial topological sort
+    res.push_back(topoSort());
+    
+    // generate all other topological sorts by permutation
+    std::vector<T>            p = res[0];
+    const unsigned int        n = size();
+    std::vector<unsigned int> loc(n);
+    unsigned int              i, k, k1;
+    T                         obj_k, obj_k1;
+    bool                      isFinal;
+    
+    for (unsigned int j = 0; j < n; ++j)
+    {
+        loc[j] = j;
+    }
+    i = 0;
+    while (i < n-1)
+    {
+        k      = loc[i];
+        k1     = k + 1;
+        obj_k  = p[k];
+        if (k1 >= n)
+        {
+            isFinal = true;
+            obj_k1  = obj_k;
+        }
+        else
+        {
+            isFinal = false;
+            obj_k1  = p[k1];
+        }
+        if (iMat[res[0][i]][obj_k1] or isFinal)
+        {
+            for (unsigned int l = k; l >= i + 1; --l)
+            {
+                p[l]   = p[l-1];
+            }
+            p[i]   = obj_k;
+            loc[i] = i;
+            i++;
+        }
+        else
+        {
+            p[k]   = obj_k1;
+            p[k1]  = obj_k;
+            loc[i] = k1;
+            i      = 0;
+            res.push_back(p);
+        }
+    }
+    
+    return res;
+}
+
+// build depedency matrix from topological sorts ///////////////////////////////
+// complexity: something like O(V^2*log(V!))
+template <typename T>
+std::map<T, std::map<T, bool>>
+makeDependencyMatrix(const std::vector<std::vector<T>> &topSort)
+{
+    std::map<T, std::map<T, bool>> m;
+    const std::vector<T>           &vList = topSort[0];
+    
+    for (auto &v1: vList)
+    for (auto &v2: vList)
+    {
+        bool dep = true;
+        
+        for (auto &t: topSort)
+        {
+            auto i1 = std::find(t.begin(), t.end(), v1);
+            auto i2 = std::find(t.begin(), t.end(), v2);
+            
+            dep = dep and (i1 - i2 > 0);
+            if (!dep) break;
+        }
+        m[v1][v2] = dep;
+    }
+    
+    return m;
+}
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_Graph_hpp_

extras/Hadrons/HadronsXmlRun.cc

@@ -0,0 +1,80 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/HadronsXmlRun.cc
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Hadrons/Application.hpp>
+
+using namespace Grid;
+using namespace QCD;
+using namespace Hadrons;
+
+int main(int argc, char *argv[])
+{
+    // parse command line
+    std::string parameterFileName, scheduleFileName = "";
+    
+    if (argc < 2)
+    {
+        std::cerr << "usage: " << argv[0] << " <parameter file> [<precomputed schedule>] [Grid options]";
+        std::cerr << std::endl;
+        std::exit(EXIT_FAILURE);
+    }
+    parameterFileName = argv[1];
+    if (argc > 2)
+    {
+        if (argv[2][0] != '-')
+        {
+            scheduleFileName = argv[2];
+        }
+    }
+    
+    // initialization
+    Grid_init(&argc, &argv);
+    HadronsLogError.Active(GridLogError.isActive());
+    HadronsLogWarning.Active(GridLogWarning.isActive());
+    HadronsLogMessage.Active(GridLogMessage.isActive());
+    HadronsLogIterative.Active(GridLogIterative.isActive());
+    HadronsLogDebug.Active(GridLogDebug.isActive());
+    LOG(Message) << "Grid initialized" << std::endl;
+    
+    // execution
+    Application application(parameterFileName);
+    
+    application.parseParameterFile(parameterFileName);
+    if (!scheduleFileName.empty())
+    {
+        application.loadSchedule(scheduleFileName);
+    }
+    application.run();
+    
+    // epilogue
+    LOG(Message) << "Grid is finalizing now" << std::endl;
+    Grid_finalize();
+    
+    return EXIT_SUCCESS;
+}

extras/Hadrons/HadronsXmlSchedule.cc

@@ -0,0 +1,72 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/HadronsXmlSchedule.cc
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Hadrons/Application.hpp>
+
+using namespace Grid;
+using namespace QCD;
+using namespace Hadrons;
+
+int main(int argc, char *argv[])
+{
+    // parse command line
+    std::string parameterFileName, scheduleFileName;
+    
+    if (argc < 3)
+    {
+        std::cerr << "usage: " << argv[0] << " <parameter file> <schedule output> [Grid options]";
+        std::cerr << std::endl;
+        std::exit(EXIT_FAILURE);
+    }
+    parameterFileName = argv[1];
+    scheduleFileName  = argv[2];
+    
+    // initialization
+    Grid_init(&argc, &argv);
+    HadronsLogError.Active(GridLogError.isActive());
+    HadronsLogWarning.Active(GridLogWarning.isActive());
+    HadronsLogMessage.Active(GridLogMessage.isActive());
+    HadronsLogIterative.Active(GridLogIterative.isActive());
+    HadronsLogDebug.Active(GridLogDebug.isActive());
+    LOG(Message) << "Grid initialized" << std::endl;
+    
+    // execution
+    Application application;
+    
+    application.parseParameterFile(parameterFileName);
+    application.schedule();
+    application.printSchedule();
+    application.saveSchedule(scheduleFileName);
+    
+    // epilogue
+    LOG(Message) << "Grid is finalizing now" << std::endl;
+    Grid_finalize();
+    
+    return EXIT_SUCCESS;
+}

extras/Hadrons/Makefile.am

@@ -0,0 +1,29 @@
+lib_LIBRARIES = libHadrons.a
+bin_PROGRAMS  = HadronsXmlRun HadronsXmlSchedule
+
+include modules.inc
+
+libHadrons_a_SOURCES = \
+    $(modules_cc)      \
+    Application.cc     \
+    Environment.cc     \
+    Global.cc          \
+    Module.cc
+libHadrons_adir = $(pkgincludedir)/Hadrons
+nobase_libHadrons_a_HEADERS = \
+	$(modules_hpp)            \
+	Application.hpp           \
+	Environment.hpp           \
+	Factory.hpp               \
+	GeneticScheduler.hpp      \
+	Global.hpp                \
+	Graph.hpp                 \
+	Module.hpp                \
+	Modules.hpp               \
+	ModuleFactory.hpp
+
+HadronsXmlRun_SOURCES = HadronsXmlRun.cc
+HadronsXmlRun_LDADD   = libHadrons.a -lGrid
+
+HadronsXmlSchedule_SOURCES = HadronsXmlSchedule.cc
+HadronsXmlSchedule_LDADD   = libHadrons.a -lGrid

extras/Hadrons/Module.cc

@@ -0,0 +1,71 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Module.cc
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Hadrons/Module.hpp>
+
+using namespace Grid;
+using namespace QCD;
+using namespace Hadrons;
+
+/******************************************************************************
+ *                       ModuleBase implementation                            *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+ModuleBase::ModuleBase(const std::string name)
+: name_(name)
+, env_(Environment::getInstance())
+{}
+
+// access //////////////////////////////////////////////////////////////////////
+std::string ModuleBase::getName(void) const
+{
+    return name_;
+}
+
+Environment & ModuleBase::env(void) const
+{
+    return env_;
+}
+
+// get factory registration name if available
+std::string ModuleBase::getRegisteredName(void)
+{
+    HADRON_ERROR("module '" + getName() + "' has a type not registered"
+                 + " in the factory");
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void ModuleBase::operator()(void)
+{
+    setup();
+    if (!env().isDryRun())
+    {
+        execute();
+    }
+}

extras/Hadrons/Module.hpp

@@ -0,0 +1,198 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Module.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_Module_hpp_
+#define Hadrons_Module_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Environment.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+// module registration macros
+#define MODULE_REGISTER(mod, base)\
+class mod: public base\
+{\
+public:\
+    typedef base Base;\
+    using Base::Base;\
+    virtual std::string getRegisteredName(void)\
+    {\
+        return std::string(#mod);\
+    }\
+};\
+class mod##ModuleRegistrar\
+{\
+public:\
+    mod##ModuleRegistrar(void)\
+    {\
+        ModuleFactory &modFac = ModuleFactory::getInstance();\
+        modFac.registerBuilder(#mod, [&](const std::string name)\
+                              {\
+                                  return std::unique_ptr<mod>(new mod(name));\
+                              });\
+    }\
+};\
+static mod##ModuleRegistrar mod##ModuleRegistrarInstance;
+
+#define MODULE_REGISTER_NS(mod, base, ns)\
+class mod: public base\
+{\
+public:\
+    typedef base Base;\
+    using Base::Base;\
+    virtual std::string getRegisteredName(void)\
+    {\
+        return std::string(#ns "::" #mod);\
+    }\
+};\
+class ns##mod##ModuleRegistrar\
+{\
+public:\
+    ns##mod##ModuleRegistrar(void)\
+    {\
+        ModuleFactory &modFac = ModuleFactory::getInstance();\
+        modFac.registerBuilder(#ns "::" #mod, [&](const std::string name)\
+                              {\
+                                  return std::unique_ptr<ns::mod>(new ns::mod(name));\
+                              });\
+    }\
+};\
+static ns##mod##ModuleRegistrar ns##mod##ModuleRegistrarInstance;
+
+#define ARG(...) __VA_ARGS__
+
+/******************************************************************************
+ *                            Module class                                    *
+ ******************************************************************************/
+// base class
+class ModuleBase
+{
+public:
+    // constructor
+    ModuleBase(const std::string name);
+    // destructor
+    virtual ~ModuleBase(void) = default;
+    // access
+    std::string getName(void) const;
+    Environment &env(void) const;
+    // get factory registration name if available
+    virtual std::string getRegisteredName(void);
+    // dependencies/products
+    virtual std::vector<std::string> getInput(void) = 0;
+    virtual std::vector<std::string> getOutput(void) = 0;
+    // parse parameters
+    virtual void parseParameters(XmlReader &reader, const std::string name) = 0;
+    virtual void saveParameters(XmlWriter &writer, const std::string name) = 0;
+    // setup
+    virtual void setup(void) {};
+    // execution
+    void operator()(void);
+    virtual void execute(void) = 0;
+private:
+    std::string name_;
+    Environment &env_;
+};
+
+// derived class, templating the parameter class
+template <typename P>
+class Module: public ModuleBase
+{
+public:
+    typedef P Par;
+public:
+    // constructor
+    Module(const std::string name);
+    // destructor
+    virtual ~Module(void) = default;
+    // parse parameters
+    virtual void parseParameters(XmlReader &reader, const std::string name);
+    virtual void saveParameters(XmlWriter &writer, const std::string name);
+    // parameter access
+    const P & par(void) const;
+    void      setPar(const P &par);
+private:
+    P par_;
+};
+
+// no parameter type
+class NoPar {};
+
+template <>
+class Module<NoPar>: public ModuleBase
+{
+public:
+    // constructor
+    Module(const std::string name): ModuleBase(name) {};
+    // destructor
+    virtual ~Module(void) = default;
+    // parse parameters (do nothing)
+    virtual void parseParameters(XmlReader &reader, const std::string name) {};
+    virtual void saveParameters(XmlWriter &writer, const std::string name)
+    {
+        push(writer, "options");
+        pop(writer);
+    };
+};
+
+/******************************************************************************
+ *                           Template implementation                          *
+ ******************************************************************************/
+template <typename P>
+Module<P>::Module(const std::string name)
+: ModuleBase(name)
+{}
+
+template <typename P>
+void Module<P>::parseParameters(XmlReader &reader, const std::string name)
+{
+    read(reader, name, par_);
+}
+
+template <typename P>
+void Module<P>::saveParameters(XmlWriter &writer, const std::string name)
+{
+    write(writer, name, par_);
+}
+
+template <typename P>
+const P & Module<P>::par(void) const
+{
+    return par_;
+}
+
+template <typename P>
+void Module<P>::setPar(const P &par)
+{
+    par_ = par;
+}
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_Module_hpp_

extras/Hadrons/ModuleFactory.hpp

@@ -0,0 +1,49 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/ModuleFactory.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_ModuleFactory_hpp_
+#define Hadrons_ModuleFactory_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Factory.hpp>
+#include <Grid/Hadrons/Module.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                            ModuleFactory                                   *
+ ******************************************************************************/
+class ModuleFactory: public Factory<ModuleBase>
+{
+    SINGLETON_DEFCTOR(ModuleFactory)
+};
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_ModuleFactory_hpp_

extras/Hadrons/Modules.hpp

@@ -0,0 +1,25 @@
+#include <Grid/Hadrons/Modules/MAction/DWF.hpp>
+#include <Grid/Hadrons/Modules/MAction/Wilson.hpp>
+#include <Grid/Hadrons/Modules/MContraction/Baryon.hpp>
+#include <Grid/Hadrons/Modules/MContraction/DiscLoop.hpp>
+#include <Grid/Hadrons/Modules/MContraction/Gamma3pt.hpp>
+#include <Grid/Hadrons/Modules/MContraction/Meson.hpp>
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonian.hpp>
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonianEye.hpp>
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.hpp>
+#include <Grid/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.hpp>
+#include <Grid/Hadrons/Modules/MFermion/GaugeProp.hpp>
+#include <Grid/Hadrons/Modules/MGauge/Load.hpp>
+#include <Grid/Hadrons/Modules/MGauge/Random.hpp>
+#include <Grid/Hadrons/Modules/MGauge/StochEm.hpp>
+#include <Grid/Hadrons/Modules/MGauge/Unit.hpp>
+#include <Grid/Hadrons/Modules/MLoop/NoiseLoop.hpp>
+#include <Grid/Hadrons/Modules/MScalar/ChargedProp.hpp>
+#include <Grid/Hadrons/Modules/MScalar/FreeProp.hpp>
+#include <Grid/Hadrons/Modules/MScalar/Scalar.hpp>
+#include <Grid/Hadrons/Modules/MSink/Point.hpp>
+#include <Grid/Hadrons/Modules/MSolver/RBPrecCG.hpp>
+#include <Grid/Hadrons/Modules/MSource/Point.hpp>
+#include <Grid/Hadrons/Modules/MSource/SeqGamma.hpp>
+#include <Grid/Hadrons/Modules/MSource/Wall.hpp>
+#include <Grid/Hadrons/Modules/MSource/Z2.hpp>

extras/Hadrons/Modules/MAction/DWF.hpp

@@ -0,0 +1,140 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MAction/DWF.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MAction_DWF_hpp_
+#define Hadrons_MAction_DWF_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                     Domain wall quark action                               *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MAction)
+
+class DWFPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(DWFPar,
+                                    std::string, gauge,
+                                    unsigned int, Ls,
+                                    double      , mass,
+                                    double      , M5,
+                                    std::string , boundary);
+};
+
+template <typename FImpl>
+class TDWF: public Module<DWFPar>
+{
+public:
+    FGS_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TDWF(const std::string name);
+    // destructor
+    virtual ~TDWF(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(DWF, TDWF<FIMPL>, MAction);
+
+/******************************************************************************
+ *                        DWF template implementation                         *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TDWF<FImpl>::TDWF(const std::string name)
+: Module<DWFPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TDWF<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().gauge};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TDWF<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TDWF<FImpl>::setup(void)
+{
+    unsigned int size;
+    
+    size = 2*env().template lattice4dSize<typename FImpl::DoubledGaugeField>();
+    env().registerObject(getName(), size, par().Ls);
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TDWF<FImpl>::execute(void)
+{
+    LOG(Message) << "Setting up domain wall fermion matrix with m= "
+                 << par().mass << ", M5= " << par().M5 << " and Ls= "
+                 << par().Ls << " using gauge field '" << par().gauge << "'"
+                 << std::endl;
+    LOG(Message) << "Fermion boundary conditions: " << par().boundary 
+                 << std::endl;
+    env().createGrid(par().Ls);
+    auto &U      = *env().template getObject<LatticeGaugeField>(par().gauge);
+    auto &g4     = *env().getGrid();
+    auto &grb4   = *env().getRbGrid();
+    auto &g5     = *env().getGrid(par().Ls);
+    auto &grb5   = *env().getRbGrid(par().Ls);
+    std::vector<Complex> boundary = strToVec<Complex>(par().boundary);
+    typename DomainWallFermion<FImpl>::ImplParams implParams(boundary);
+    FMat *fMatPt = new DomainWallFermion<FImpl>(U, g5, grb5, g4, grb4,
+                                                par().mass, par().M5,
+                                                implParams);
+    env().setObject(getName(), fMatPt);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MAction_DWF_hpp_

extras/Hadrons/Modules/MAction/Wilson.hpp

@@ -0,0 +1,132 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MAction/Wilson.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MAction_Wilson_hpp_
+#define Hadrons_MAction_Wilson_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                            TWilson quark action                            *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MAction)
+
+class WilsonPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(WilsonPar,
+                                    std::string, gauge,
+                                    double     , mass,
+                                    std::string, boundary);
+};
+
+template <typename FImpl>
+class TWilson: public Module<WilsonPar>
+{
+public:
+    FGS_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TWilson(const std::string name);
+    // destructor
+    virtual ~TWilson(void) = default;
+    // dependencies/products
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(Wilson, TWilson<FIMPL>, MAction);
+
+/******************************************************************************
+ *                     TWilson template implementation                        *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TWilson<FImpl>::TWilson(const std::string name)
+: Module<WilsonPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TWilson<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().gauge};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TWilson<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TWilson<FImpl>::setup(void)
+{
+    unsigned int size;
+    
+    size = 2*env().template lattice4dSize<typename FImpl::DoubledGaugeField>();
+    env().registerObject(getName(), size);
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TWilson<FImpl>::execute()
+{
+    LOG(Message) << "Setting up TWilson fermion matrix with m= " << par().mass
+                 << " using gauge field '" << par().gauge << "'" << std::endl;
+    LOG(Message) << "Fermion boundary conditions: " << par().boundary 
+                 << std::endl;
+    auto &U      = *env().template getObject<LatticeGaugeField>(par().gauge);
+    auto &grid   = *env().getGrid();
+    auto &gridRb = *env().getRbGrid();
+    std::vector<Complex> boundary = strToVec<Complex>(par().boundary);
+    typename WilsonFermion<FImpl>::ImplParams implParams(boundary);
+    FMat *fMatPt = new WilsonFermion<FImpl>(U, grid, gridRb, par().mass,
+                                            implParams);
+    env().setObject(getName(), fMatPt);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_Wilson_hpp_

extras/Hadrons/Modules/MContraction/Baryon.hpp

@@ -0,0 +1,131 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/Baryon.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MContraction_Baryon_hpp_
+#define Hadrons_MContraction_Baryon_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                               Baryon                                       *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+class BaryonPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(BaryonPar,
+                                    std::string, q1,
+                                    std::string, q2,
+                                    std::string, q3,
+                                    std::string, output);
+};
+
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+class TBaryon: public Module<BaryonPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl1, 1);
+    FERM_TYPE_ALIASES(FImpl2, 2);
+    FERM_TYPE_ALIASES(FImpl3, 3);
+    class Result: Serializable
+    {
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
+                                        std::vector<std::vector<std::vector<Complex>>>, corr);
+    };
+public:
+    // constructor
+    TBaryon(const std::string name);
+    // destructor
+    virtual ~TBaryon(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(Baryon, ARG(TBaryon<FIMPL, FIMPL, FIMPL>), MContraction);
+
+/******************************************************************************
+ *                         TBaryon implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+TBaryon<FImpl1, FImpl2, FImpl3>::TBaryon(const std::string name)
+: Module<BaryonPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+std::vector<std::string> TBaryon<FImpl1, FImpl2, FImpl3>::getInput(void)
+{
+    std::vector<std::string> input = {par().q1, par().q2, par().q3};
+    
+    return input;
+}
+
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+std::vector<std::string> TBaryon<FImpl1, FImpl2, FImpl3>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+void TBaryon<FImpl1, FImpl2, FImpl3>::execute(void)
+{
+    LOG(Message) << "Computing baryon contractions '" << getName() << "' using"
+                 << " quarks '" << par().q1 << "', '" << par().q2 << "', and '"
+                 << par().q3 << "'" << std::endl;
+    
+    CorrWriter             writer(par().output);
+    PropagatorField1      &q1 = *env().template getObject<PropagatorField1>(par().q1);
+    PropagatorField2      &q2 = *env().template getObject<PropagatorField2>(par().q2);
+    PropagatorField3      &q3 = *env().template getObject<PropagatorField3>(par().q2);
+    LatticeComplex        c(env().getGrid());
+    Result                result;
+    
+    // FIXME: do contractions
+    
+    // write(writer, "meson", result);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_Baryon_hpp_

extras/Hadrons/Modules/MContraction/DiscLoop.hpp

@@ -0,0 +1,144 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/DiscLoop.hpp
+
+Copyright (C) 2017
+
+Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MContraction_DiscLoop_hpp_
+#define Hadrons_MContraction_DiscLoop_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                                DiscLoop                                    *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+class DiscLoopPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(DiscLoopPar,
+                                    std::string,    q_loop,
+                                    Gamma::Algebra, gamma,
+                                    std::string,    output);
+};
+
+template <typename FImpl>
+class TDiscLoop: public Module<DiscLoopPar>
+{
+    FERM_TYPE_ALIASES(FImpl,);
+    class Result: Serializable
+    {
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
+                                        Gamma::Algebra, gamma,
+                                        std::vector<Complex>, corr);
+    };
+public:
+    // constructor
+    TDiscLoop(const std::string name);
+    // destructor
+    virtual ~TDiscLoop(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(DiscLoop, TDiscLoop<FIMPL>, MContraction);
+
+/******************************************************************************
+ *                       TDiscLoop implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TDiscLoop<FImpl>::TDiscLoop(const std::string name)
+: Module<DiscLoopPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TDiscLoop<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().q_loop};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TDiscLoop<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TDiscLoop<FImpl>::setup(void)
+{
+    
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TDiscLoop<FImpl>::execute(void)
+{
+    LOG(Message) << "Computing disconnected loop contraction '" << getName() 
+                 << "' using '" << par().q_loop << "' with " << par().gamma 
+                 << " insertion." << std::endl;
+
+    CorrWriter            writer(par().output);
+    PropagatorField       &q_loop = *env().template getObject<PropagatorField>(par().q_loop);
+    LatticeComplex        c(env().getGrid());
+    Gamma                 gamma(par().gamma);
+    std::vector<TComplex> buf;
+    Result                result;
+
+    c = trace(gamma*q_loop);
+    sliceSum(c, buf, Tp);
+
+    result.gamma = par().gamma;
+    result.corr.resize(buf.size());
+    for (unsigned int t = 0; t < buf.size(); ++t)
+    {
+        result.corr[t] = TensorRemove(buf[t]);
+    }
+
+    write(writer, "disc", result);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_DiscLoop_hpp_

extras/Hadrons/Modules/MContraction/Gamma3pt.hpp

@@ -0,0 +1,170 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/Gamma3pt.hpp
+
+Copyright (C) 2017
+
+Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MContraction_Gamma3pt_hpp_
+#define Hadrons_MContraction_Gamma3pt_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/*
+ * 3pt contraction with gamma matrix insertion.
+ *
+ * Schematic:
+ *
+ *             q2           q3
+ *        /----<------*------<----¬
+ *       /          gamma          \
+ *      /                           \
+ *   i *                            * f
+ *      \                          /
+ *       \                        /
+ *        \----------->----------/
+ *                   q1
+ *
+ *      trace(g5*q1*adj(q2)*g5*gamma*q3)
+ */
+
+/******************************************************************************
+ *                               Gamma3pt                                     *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+class Gamma3ptPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(Gamma3ptPar,
+                                    std::string,    q1,
+                                    std::string,    q2,
+                                    std::string,    q3,
+                                    Gamma::Algebra, gamma,
+                                    std::string,    output);
+};
+
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+class TGamma3pt: public Module<Gamma3ptPar>
+{
+    FERM_TYPE_ALIASES(FImpl1, 1);
+    FERM_TYPE_ALIASES(FImpl2, 2);
+    FERM_TYPE_ALIASES(FImpl3, 3);
+    class Result: Serializable
+    {
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
+                                        Gamma::Algebra, gamma,
+                                        std::vector<Complex>, corr);
+    };
+public:
+    // constructor
+    TGamma3pt(const std::string name);
+    // destructor
+    virtual ~TGamma3pt(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(Gamma3pt, ARG(TGamma3pt<FIMPL, FIMPL, FIMPL>), MContraction);
+
+/******************************************************************************
+ *                       TGamma3pt implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+TGamma3pt<FImpl1, FImpl2, FImpl3>::TGamma3pt(const std::string name)
+: Module<Gamma3ptPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+std::vector<std::string> TGamma3pt<FImpl1, FImpl2, FImpl3>::getInput(void)
+{
+    std::vector<std::string> in = {par().q1, par().q2, par().q3};
+    
+    return in;
+}
+
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+std::vector<std::string> TGamma3pt<FImpl1, FImpl2, FImpl3>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+void TGamma3pt<FImpl1, FImpl2, FImpl3>::setup(void)
+{
+    
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+void TGamma3pt<FImpl1, FImpl2, FImpl3>::execute(void)
+{
+    LOG(Message) << "Computing 3pt contractions '" << getName() << "' using"
+                 << " quarks '" << par().q1 << "', '" << par().q2 << "' and '"
+                 << par().q3 << "', with " << par().gamma << " insertion." 
+                 << std::endl;
+
+    CorrWriter            writer(par().output);
+    PropagatorField1      &q1 = *env().template getObject<PropagatorField1>(par().q1);
+    PropagatorField2      &q2 = *env().template getObject<PropagatorField2>(par().q2);
+    PropagatorField3      &q3 = *env().template getObject<PropagatorField3>(par().q3);
+    LatticeComplex        c(env().getGrid());
+    Gamma                 g5(Gamma::Algebra::Gamma5);
+    Gamma                 gamma(par().gamma);
+    std::vector<TComplex> buf;
+    Result                result;
+
+    c = trace(g5*q1*adj(q2)*(g5*gamma)*q3);
+    sliceSum(c, buf, Tp);
+
+    result.gamma = par().gamma;
+    result.corr.resize(buf.size());
+    for (unsigned int t = 0; t < buf.size(); ++t)
+    {
+        result.corr[t] = TensorRemove(buf[t]);
+    }
+
+    write(writer, "gamma3pt", result);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_Gamma3pt_hpp_

extras/Hadrons/Modules/MContraction/Meson.hpp

@@ -0,0 +1,244 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/Meson.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+Copyright (C) 2017
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+        Andrew Lawson    <andrew.lawson1991@gmail.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MContraction_Meson_hpp_
+#define Hadrons_MContraction_Meson_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/*
+ 
+ Meson contractions
+ -----------------------------
+ 
+ * options:
+ - q1: input propagator 1 (string)
+ - q2: input propagator 2 (string)
+ - gammas: gamma products to insert at sink & source, pairs of gamma matrices 
+           (space-separated strings) in angled brackets (i.e. <g_sink g_src>),
+           in a sequence (e.g. "<Gamma5 Gamma5><Gamma5 GammaT>").
+
+           Special values: "all" - perform all possible contractions.
+ - mom: momentum insertion, space-separated float sequence (e.g ".1 .2 1. 0."),
+        given as multiples of (2*pi) / L.
+*/
+
+/******************************************************************************
+ *                                TMeson                                       *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+typedef std::pair<Gamma::Algebra, Gamma::Algebra> GammaPair;
+
+class MesonPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(MesonPar,
+                                    std::string, q1,
+                                    std::string, q2,
+                                    std::string, gammas,
+                                    std::string, sink,
+                                    std::string, output);
+};
+
+template <typename FImpl1, typename FImpl2>
+class TMeson: public Module<MesonPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl1, 1);
+    FERM_TYPE_ALIASES(FImpl2, 2);
+    FERM_TYPE_ALIASES(ScalarImplCR, Scalar);
+    SINK_TYPE_ALIASES(Scalar);
+    class Result: Serializable
+    {
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
+                                        Gamma::Algebra, gamma_snk,
+                                        Gamma::Algebra, gamma_src,
+                                        std::vector<Complex>, corr);
+    };
+public:
+    // constructor
+    TMeson(const std::string name);
+    // destructor
+    virtual ~TMeson(void) = default;
+    // dependencies/products
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    virtual void parseGammaString(std::vector<GammaPair> &gammaList);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(Meson, ARG(TMeson<FIMPL, FIMPL>), MContraction);
+
+/******************************************************************************
+ *                           TMeson implementation                            *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2>
+TMeson<FImpl1, FImpl2>::TMeson(const std::string name)
+: Module<MesonPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2>
+std::vector<std::string> TMeson<FImpl1, FImpl2>::getInput(void)
+{
+    std::vector<std::string> input = {par().q1, par().q2, par().sink};
+    
+    return input;
+}
+
+template <typename FImpl1, typename FImpl2>
+std::vector<std::string> TMeson<FImpl1, FImpl2>::getOutput(void)
+{
+    std::vector<std::string> output = {getName()};
+    
+    return output;
+}
+
+template <typename FImpl1, typename FImpl2>
+void TMeson<FImpl1, FImpl2>::parseGammaString(std::vector<GammaPair> &gammaList)
+{
+    gammaList.clear();
+    // Determine gamma matrices to insert at source/sink.
+    if (par().gammas.compare("all") == 0)
+    {
+        // Do all contractions.
+        for (unsigned int i = 1; i < Gamma::nGamma; i += 2)
+        {
+            for (unsigned int j = 1; j < Gamma::nGamma; j += 2)
+            {
+                gammaList.push_back(std::make_pair((Gamma::Algebra)i, 
+                                                   (Gamma::Algebra)j));
+            }
+        }
+    }
+    else
+    {
+        // Parse individual contractions from input string.
+        gammaList = strToVec<GammaPair>(par().gammas);
+    }
+}
+
+
+// execution ///////////////////////////////////////////////////////////////////
+#define mesonConnected(q1, q2, gSnk, gSrc) \
+(g5*(gSnk))*(q1)*(adj(gSrc)*g5)*adj(q2)
+
+template <typename FImpl1, typename FImpl2>
+void TMeson<FImpl1, FImpl2>::execute(void)
+{
+    LOG(Message) << "Computing meson contractions '" << getName() << "' using"
+                 << " quarks '" << par().q1 << "' and '" << par().q2 << "'"
+                 << std::endl;
+    
+    CorrWriter             writer(par().output);
+    std::vector<TComplex>  buf;
+    std::vector<Result>    result;
+    Gamma                  g5(Gamma::Algebra::Gamma5);
+    std::vector<GammaPair> gammaList;
+    int                    nt = env().getDim(Tp);
+    
+    parseGammaString(gammaList);
+    result.resize(gammaList.size());
+    for (unsigned int i = 0; i < result.size(); ++i)
+    {
+        result[i].gamma_snk = gammaList[i].first;
+        result[i].gamma_src = gammaList[i].second;
+        result[i].corr.resize(nt);
+    }
+    if (env().template isObjectOfType<SlicedPropagator1>(par().q1) and
+        env().template isObjectOfType<SlicedPropagator2>(par().q2))
+    {
+        SlicedPropagator1 &q1 = *env().template getObject<SlicedPropagator1>(par().q1);
+        SlicedPropagator2 &q2 = *env().template getObject<SlicedPropagator2>(par().q2);
+        
+        LOG(Message) << "(propagator already sinked)" << std::endl;
+        for (unsigned int i = 0; i < result.size(); ++i)
+        {
+            Gamma gSnk(gammaList[i].first);
+            Gamma gSrc(gammaList[i].second);
+            
+            for (unsigned int t = 0; t < buf.size(); ++t)
+            {
+                result[i].corr[t] = TensorRemove(trace(mesonConnected(q1[t], q2[t], gSnk, gSrc)));
+            }
+        }
+    }
+    else
+    {
+        PropagatorField1 &q1   = *env().template getObject<PropagatorField1>(par().q1);
+        PropagatorField2 &q2   = *env().template getObject<PropagatorField2>(par().q2);
+        LatticeComplex   c(env().getGrid());
+        
+        LOG(Message) << "(using sink '" << par().sink << "')" << std::endl;
+        for (unsigned int i = 0; i < result.size(); ++i)
+        {
+            Gamma       gSnk(gammaList[i].first);
+            Gamma       gSrc(gammaList[i].second);
+            std::string ns;
+                
+            ns = env().getModuleNamespace(env().getObjectModule(par().sink));
+            if (ns == "MSource")
+            {
+                PropagatorField1 &sink =
+                    *env().template getObject<PropagatorField1>(par().sink);
+                
+                c = trace(mesonConnected(q1, q2, gSnk, gSrc)*sink);
+                sliceSum(c, buf, Tp);
+            }
+            else if (ns == "MSink")
+            {
+                SinkFnScalar &sink = *env().template getObject<SinkFnScalar>(par().sink);
+                
+                c   = trace(mesonConnected(q1, q2, gSnk, gSrc));
+                buf = sink(c);
+            }
+            for (unsigned int t = 0; t < buf.size(); ++t)
+            {
+                result[i].corr[t] = TensorRemove(buf[t]);
+            }
+        }
+    }
+    write(writer, "meson", result);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_Meson_hpp_

extras/Hadrons/Modules/MContraction/WeakHamiltonian.hpp

@@ -0,0 +1,114 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonian.hpp
+
+Copyright (C) 2017
+
+Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MContraction_WeakHamiltonian_hpp_
+#define Hadrons_MContraction_WeakHamiltonian_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         WeakHamiltonian                                    *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+/*******************************************************************************
+ * Utilities for contractions involving the Weak Hamiltonian.
+ ******************************************************************************/
+//// Sum and store correlator.
+#define MAKE_DIAG(exp, buf, res, n)\
+sliceSum(exp, buf, Tp);\
+res.name = (n);\
+res.corr.resize(buf.size());\
+for (unsigned int t = 0; t < buf.size(); ++t)\
+{\
+    res.corr[t] = TensorRemove(buf[t]);\
+}
+
+//// Contraction of mu index: use 'mu' variable in exp.
+#define SUM_MU(buf,exp)\
+buf = zero;\
+for (unsigned int mu = 0; mu < ndim; ++mu)\
+{\
+    buf += exp;\
+}
+
+enum 
+{
+  i_V = 0,
+  i_A = 1,
+  n_i = 2
+};
+
+class WeakHamiltonianPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(WeakHamiltonianPar,
+                                    std::string, q1,
+                                    std::string, q2,
+                                    std::string, q3,
+                                    std::string, q4,
+                                    std::string, output);
+};
+
+#define MAKE_WEAK_MODULE(modname)\
+class T##modname: public Module<WeakHamiltonianPar>\
+{\
+public:\
+    FERM_TYPE_ALIASES(FIMPL,)\
+    class Result: Serializable\
+    {\
+    public:\
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,\
+                                        std::string, name,\
+                                        std::vector<Complex>, corr);\
+    };\
+public:\
+    /* constructor */ \
+    T##modname(const std::string name);\
+    /* destructor */ \
+    virtual ~T##modname(void) = default;\
+    /* dependency relation */ \
+    virtual std::vector<std::string> getInput(void);\
+    virtual std::vector<std::string> getOutput(void);\
+    /* setup */ \
+    virtual void setup(void);\
+    /* execution */ \
+    virtual void execute(void);\
+    std::vector<std::string> VA_label = {"V", "A"};\
+};\
+MODULE_REGISTER_NS(modname, T##modname, MContraction);
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_WeakHamiltonian_hpp_

extras/Hadrons/Modules/MContraction/WeakHamiltonianEye.cc

@@ -0,0 +1,137 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonianEye.cc
+
+Copyright (C) 2017
+
+Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonianEye.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MContraction;
+
+/*
+ * Weak Hamiltonian current-current contractions, Eye-type.
+ * 
+ * These contractions are generated by the Q1 and Q2 operators in the physical
+ * basis (see e.g. Fig 3 of arXiv:1507.03094).
+ * 
+ * Schematics:        q4                 |                  
+ *                  /-<-¬                |                             
+ *                 /     \               |             q2           q3
+ *                 \     /               |        /----<------*------<----¬                        
+ *            q2    \   /    q3          |       /          /-*-¬          \
+ *       /-----<-----* *-----<----¬      |      /          /     \          \
+ *    i *            H_W           * f   |   i *           \     /  q4      * f
+ *       \                        /      |      \           \->-/          /   
+ *        \                      /       |       \                        /       
+ *         \---------->---------/        |        \----------->----------/        
+ *                   q1                  |                   q1                  
+ *                                       |
+ *                Saucer (S)             |                  Eye (E)
+ * 
+ * S: trace(q3*g5*q1*adj(q2)*g5*gL[mu][p_1]*q4*gL[mu][p_2])
+ * E: trace(q3*g5*q1*adj(q2)*g5*gL[mu][p_1])*trace(q4*gL[mu][p_2])
+ */
+
+/******************************************************************************
+ *                  TWeakHamiltonianEye implementation                        *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TWeakHamiltonianEye::TWeakHamiltonianEye(const std::string name)
+: Module<WeakHamiltonianPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TWeakHamiltonianEye::getInput(void)
+{
+    std::vector<std::string> in = {par().q1, par().q2, par().q3, par().q4};
+    
+    return in;
+}
+
+std::vector<std::string> TWeakHamiltonianEye::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TWeakHamiltonianEye::setup(void)
+{
+
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TWeakHamiltonianEye::execute(void)
+{
+    LOG(Message) << "Computing Weak Hamiltonian (Eye type) contractions '" 
+                 << getName() << "' using quarks '" << par().q1 << "', '" 
+                 << par().q2 << ", '" << par().q3 << "' and '" << par().q4 
+                 << "'." << std::endl;
+
+    CorrWriter             writer(par().output);
+    PropagatorField &q1 = *env().template getObject<PropagatorField>(par().q1);
+    PropagatorField &q2 = *env().template getObject<PropagatorField>(par().q2);
+    PropagatorField &q3 = *env().template getObject<PropagatorField>(par().q3);
+    PropagatorField &q4 = *env().template getObject<PropagatorField>(par().q4);
+    Gamma g5            = Gamma(Gamma::Algebra::Gamma5);
+    LatticeComplex        expbuf(env().getGrid());
+    std::vector<TComplex> corrbuf;
+    std::vector<Result>   result(n_eye_diag);
+    unsigned int ndim   = env().getNd();
+
+    PropagatorField              tmp1(env().getGrid());
+    LatticeComplex               tmp2(env().getGrid());
+    std::vector<PropagatorField> S_body(ndim, tmp1);
+    std::vector<PropagatorField> S_loop(ndim, tmp1);
+    std::vector<LatticeComplex>  E_body(ndim, tmp2);
+    std::vector<LatticeComplex>  E_loop(ndim, tmp2);
+
+    // Setup for S-type contractions.
+    for (int mu = 0; mu < ndim; ++mu)
+    {
+        S_body[mu] = MAKE_SE_BODY(q1, q2, q3, GammaL(Gamma::gmu[mu]));
+        S_loop[mu] = MAKE_SE_LOOP(q4, GammaL(Gamma::gmu[mu]));
+    }
+
+    // Perform S-type contractions.    
+    SUM_MU(expbuf, trace(S_body[mu]*S_loop[mu]))
+    MAKE_DIAG(expbuf, corrbuf, result[S_diag], "HW_S")
+
+    // Recycle sub-expressions for E-type contractions.
+    for (unsigned int mu = 0; mu < ndim; ++mu)
+    {
+        E_body[mu] = trace(S_body[mu]);
+        E_loop[mu] = trace(S_loop[mu]);
+    }
+
+    // Perform E-type contractions.
+    SUM_MU(expbuf, E_body[mu]*E_loop[mu])
+    MAKE_DIAG(expbuf, corrbuf, result[E_diag], "HW_E")
+
+    write(writer, "HW_Eye", result);
+}

extras/Hadrons/Modules/MContraction/WeakHamiltonianEye.hpp

@@ -0,0 +1,58 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonianEye.hpp
+
+Copyright (C) 2017
+
+Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MContraction_WeakHamiltonianEye_hpp_
+#define Hadrons_MContraction_WeakHamiltonianEye_hpp_
+
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonian.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         WeakHamiltonianEye                                 *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+enum
+{
+    S_diag = 0,
+    E_diag = 1,
+    n_eye_diag = 2
+};
+
+// Saucer and Eye subdiagram contractions.
+#define MAKE_SE_BODY(Q_1, Q_2, Q_3, gamma) (Q_3*g5*Q_1*adj(Q_2)*g5*gamma)
+#define MAKE_SE_LOOP(Q_loop, gamma) (Q_loop*gamma)
+
+MAKE_WEAK_MODULE(WeakHamiltonianEye)
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_WeakHamiltonianEye_hpp_

extras/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.cc

@@ -0,0 +1,139 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.cc
+
+Copyright (C) 2017
+
+Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MContraction;
+
+/*
+ * Weak Hamiltonian current-current contractions, Non-Eye-type.
+ * 
+ * These contractions are generated by the Q1 and Q2 operators in the physical
+ * basis (see e.g. Fig 3 of arXiv:1507.03094).
+ * 
+ * Schematic:     
+ *            q2             q3          |           q2              q3
+ *          /--<--¬       /--<--¬        |        /--<--¬         /--<--¬       
+ *         /       \     /       \       |       /       \       /       \      
+ *        /         \   /         \      |      /         \     /         \     
+ *       /           \ /           \     |     /           \   /           \    
+ *    i *             * H_W         *  f |  i *             * * H_W         * f 
+ *      \             *             |    |     \           /   \           /
+ *       \           / \           /     |      \         /     \         /    
+ *        \         /   \         /      |       \       /       \       /  
+ *         \       /     \       /       |        \-->--/         \-->--/      
+ *          \-->--/       \-->--/        |          q1               q4 
+ *            q1             q4          |
+ *                Connected (C)          |                 Wing (W)
+ *
+ * C: trace(q1*adj(q2)*g5*gL[mu]*q3*adj(q4)*g5*gL[mu])
+ * W: trace(q1*adj(q2)*g5*gL[mu])*trace(q3*adj(q4)*g5*gL[mu])
+ * 
+ */
+
+/******************************************************************************
+ *                  TWeakHamiltonianNonEye implementation                     *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TWeakHamiltonianNonEye::TWeakHamiltonianNonEye(const std::string name)
+: Module<WeakHamiltonianPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TWeakHamiltonianNonEye::getInput(void)
+{
+    std::vector<std::string> in = {par().q1, par().q2, par().q3, par().q4};
+    
+    return in;
+}
+
+std::vector<std::string> TWeakHamiltonianNonEye::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TWeakHamiltonianNonEye::setup(void)
+{
+
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TWeakHamiltonianNonEye::execute(void)
+{
+    LOG(Message) << "Computing Weak Hamiltonian (Non-Eye type) contractions '" 
+                 << getName() << "' using quarks '" << par().q1 << "', '" 
+                 << par().q2 << ", '" << par().q3 << "' and '" << par().q4 
+                 << "'." << std::endl;
+    
+    CorrWriter             writer(par().output);
+    PropagatorField &q1 = *env().template getObject<PropagatorField>(par().q1);
+    PropagatorField &q2 = *env().template getObject<PropagatorField>(par().q2);
+    PropagatorField &q3 = *env().template getObject<PropagatorField>(par().q3);
+    PropagatorField &q4 = *env().template getObject<PropagatorField>(par().q4);
+    Gamma g5            = Gamma(Gamma::Algebra::Gamma5);
+    LatticeComplex        expbuf(env().getGrid());
+    std::vector<TComplex> corrbuf;
+    std::vector<Result>   result(n_noneye_diag); 
+    unsigned int ndim   = env().getNd();
+
+    PropagatorField              tmp1(env().getGrid());
+    LatticeComplex               tmp2(env().getGrid());
+    std::vector<PropagatorField> C_i_side_loop(ndim, tmp1);
+    std::vector<PropagatorField> C_f_side_loop(ndim, tmp1);
+    std::vector<LatticeComplex>  W_i_side_loop(ndim, tmp2);
+    std::vector<LatticeComplex>  W_f_side_loop(ndim, tmp2);
+
+    // Setup for C-type contractions.
+    for (int mu = 0; mu < ndim; ++mu)
+    {
+        C_i_side_loop[mu] = MAKE_CW_SUBDIAG(q1, q2, GammaL(Gamma::gmu[mu]));
+        C_f_side_loop[mu] = MAKE_CW_SUBDIAG(q3, q4, GammaL(Gamma::gmu[mu]));
+    }
+
+    // Perform C-type contractions.    
+    SUM_MU(expbuf, trace(C_i_side_loop[mu]*C_f_side_loop[mu]))
+    MAKE_DIAG(expbuf, corrbuf, result[C_diag], "HW_C")
+
+    // Recycle sub-expressions for W-type contractions.
+    for (unsigned int mu = 0; mu < ndim; ++mu)
+    {
+        W_i_side_loop[mu] = trace(C_i_side_loop[mu]);
+        W_f_side_loop[mu] = trace(C_f_side_loop[mu]);
+    }
+
+    // Perform W-type contractions.
+    SUM_MU(expbuf, W_i_side_loop[mu]*W_f_side_loop[mu])
+    MAKE_DIAG(expbuf, corrbuf, result[W_diag], "HW_W")
+
+    write(writer, "HW_NonEye", result);
+}

extras/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.hpp

@@ -0,0 +1,57 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.hpp
+
+Copyright (C) 2017
+
+Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MContraction_WeakHamiltonianNonEye_hpp_
+#define Hadrons_MContraction_WeakHamiltonianNonEye_hpp_
+
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonian.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         WeakHamiltonianNonEye                              *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+enum
+{
+    W_diag = 0,
+    C_diag = 1,
+    n_noneye_diag = 2
+};
+
+// Wing and Connected subdiagram contractions
+#define MAKE_CW_SUBDIAG(Q_1, Q_2, gamma) (Q_1*adj(Q_2)*g5*gamma)
+
+MAKE_WEAK_MODULE(WeakHamiltonianNonEye)
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_WeakHamiltonianNonEye_hpp_

extras/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.cc

@@ -0,0 +1,135 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.cc
+
+Copyright (C) 2017
+
+Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MContraction;
+
+/*
+ * Weak Hamiltonian + current contractions, disconnected topology for neutral 
+ * mesons.
+ * 
+ * These contractions are generated by operators Q_1,...,10 of the dS=1 Weak
+ * Hamiltonian in the physical basis and an additional current J (see e.g. 
+ * Fig 11 of arXiv:1507.03094).
+ * 
+ * Schematic:
+ *                        
+ *           q2          q4             q3
+ *       /--<--¬     /---<--¬       /---<--¬
+ *     /         \ /         \     /        \
+ *  i *           * H_W      |  J *          * f
+ *     \         / \         /     \        /
+ *      \--->---/   \-------/       \------/
+ *          q1 
+ * 
+ * options
+ * - q1: input propagator 1 (string)
+ * - q2: input propagator 2 (string)
+ * - q3: input propagator 3 (string), assumed to be sequential propagator 
+ * - q4: input propagator 4 (string), assumed to be a loop
+ * 
+ * type 1: trace(q1*adj(q2)*g5*gL[mu])*trace(loop*gL[mu])*trace(q3*g5)
+ * type 2: trace(q1*adj(q2)*g5*gL[mu]*loop*gL[mu])*trace(q3*g5)
+ */
+
+/*******************************************************************************
+ *                  TWeakNeutral4ptDisc implementation                         *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TWeakNeutral4ptDisc::TWeakNeutral4ptDisc(const std::string name)
+: Module<WeakHamiltonianPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TWeakNeutral4ptDisc::getInput(void)
+{
+    std::vector<std::string> in = {par().q1, par().q2, par().q3, par().q4};
+    
+    return in;
+}
+
+std::vector<std::string> TWeakNeutral4ptDisc::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TWeakNeutral4ptDisc::setup(void)
+{
+
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TWeakNeutral4ptDisc::execute(void)
+{
+    LOG(Message) << "Computing Weak Hamiltonian neutral disconnected contractions '" 
+                 << getName() << "' using quarks '" << par().q1 << "', '" 
+                 << par().q2 << ", '" << par().q3 << "' and '" << par().q4 
+                 << "'." << std::endl;
+
+    CorrWriter             writer(par().output);
+    PropagatorField &q1 = *env().template getObject<PropagatorField>(par().q1);
+    PropagatorField &q2 = *env().template getObject<PropagatorField>(par().q2);
+    PropagatorField &q3 = *env().template getObject<PropagatorField>(par().q3);
+    PropagatorField &q4 = *env().template getObject<PropagatorField>(par().q4);
+    Gamma g5            = Gamma(Gamma::Algebra::Gamma5);
+    LatticeComplex        expbuf(env().getGrid());
+    std::vector<TComplex> corrbuf;
+    std::vector<Result>   result(n_neut_disc_diag);
+    unsigned int ndim   = env().getNd();
+
+    PropagatorField              tmp(env().getGrid());
+    std::vector<PropagatorField> meson(ndim, tmp);
+    std::vector<PropagatorField> loop(ndim, tmp);
+    LatticeComplex               curr(env().getGrid());
+
+    // Setup for type 1 contractions.
+    for (int mu = 0; mu < ndim; ++mu)
+    {
+        meson[mu] = MAKE_DISC_MESON(q1, q2, GammaL(Gamma::gmu[mu]));
+        loop[mu] = MAKE_DISC_LOOP(q4, GammaL(Gamma::gmu[mu]));
+    }
+    curr = MAKE_DISC_CURR(q3, GammaL(Gamma::Algebra::Gamma5));
+
+    // Perform type 1 contractions.    
+    SUM_MU(expbuf, trace(meson[mu]*loop[mu]))
+    expbuf *= curr;
+    MAKE_DIAG(expbuf, corrbuf, result[neut_disc_1_diag], "HW_disc0_1")
+
+    // Perform type 2 contractions.
+    SUM_MU(expbuf, trace(meson[mu])*trace(loop[mu]))
+    expbuf *= curr;
+    MAKE_DIAG(expbuf, corrbuf, result[neut_disc_2_diag], "HW_disc0_2")
+
+    write(writer, "HW_disc0", result);
+}

extras/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.hpp

@@ -0,0 +1,59 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.hpp
+
+Copyright (C) 2017
+
+Author: Andrew Lawson    <andrew.lawson1991@gmail.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MContraction_WeakNeutral4ptDisc_hpp_
+#define Hadrons_MContraction_WeakNeutral4ptDisc_hpp_
+
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonian.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         WeakNeutral4ptDisc                                 *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+enum
+{
+    neut_disc_1_diag = 0,
+    neut_disc_2_diag = 1,
+    n_neut_disc_diag = 2
+};
+
+// Neutral 4pt disconnected subdiagram contractions.
+#define MAKE_DISC_MESON(Q_1, Q_2, gamma) (Q_1*adj(Q_2)*g5*gamma)
+#define MAKE_DISC_LOOP(Q_LOOP, gamma) (Q_LOOP*gamma)
+#define MAKE_DISC_CURR(Q_c, gamma) (trace(Q_c*gamma))
+
+MAKE_WEAK_MODULE(WeakNeutral4ptDisc)
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_WeakNeutral4ptDisc_hpp_

extras/Hadrons/Modules/MFermion/GaugeProp.hpp

@@ -0,0 +1,160 @@
+#ifndef Hadrons_MFermion_GaugeProp_hpp_
+#define Hadrons_MFermion_GaugeProp_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                                GaugeProp                                   *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MFermion)
+
+class GaugePropPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(GaugePropPar,
+                                    std::string, source,
+                                    std::string, solver);
+};
+
+template <typename FImpl>
+class TGaugeProp: public Module<GaugePropPar>
+{
+public:
+    FGS_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TGaugeProp(const std::string name);
+    // destructor
+    virtual ~TGaugeProp(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+private:
+    unsigned int Ls_;
+    SolverFn     *solver_{nullptr};
+};
+
+MODULE_REGISTER_NS(GaugeProp, TGaugeProp<FIMPL>, MFermion);
+
+/******************************************************************************
+ *                      TGaugeProp implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TGaugeProp<FImpl>::TGaugeProp(const std::string name)
+: Module<GaugePropPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TGaugeProp<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().source, par().solver};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TGaugeProp<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName(), getName() + "_5d"};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TGaugeProp<FImpl>::setup(void)
+{
+    Ls_ = env().getObjectLs(par().solver);
+    env().template registerLattice<PropagatorField>(getName());
+    if (Ls_ > 1)
+    {
+        env().template registerLattice<PropagatorField>(getName() + "_5d", Ls_);
+    }
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TGaugeProp<FImpl>::execute(void)
+{
+    LOG(Message) << "Computing quark propagator '" << getName() << "'"
+    << std::endl;
+    
+    FermionField    source(env().getGrid(Ls_)), sol(env().getGrid(Ls_)),
+    tmp(env().getGrid());
+    std::string     propName = (Ls_ == 1) ? getName() : (getName() + "_5d");
+    PropagatorField &prop    = *env().template createLattice<PropagatorField>(propName);
+    PropagatorField &fullSrc = *env().template getObject<PropagatorField>(par().source);
+    SolverFn        &solver  = *env().template getObject<SolverFn>(par().solver);
+    if (Ls_ > 1)
+    {
+        env().template createLattice<PropagatorField>(getName());
+    }
+    
+    LOG(Message) << "Inverting using solver '" << par().solver
+    << "' on source '" << par().source << "'" << std::endl;
+    for (unsigned int s = 0; s < Ns; ++s)
+    for (unsigned int c = 0; c < Nc; ++c)
+    {
+        LOG(Message) << "Inversion for spin= " << s << ", color= " << c
+        << std::endl;
+        // source conversion for 4D sources
+        if (!env().isObject5d(par().source))
+        {
+            if (Ls_ == 1)
+            {
+                PropToFerm(source, fullSrc, s, c);
+            }
+            else
+            {
+                source = zero;
+                PropToFerm(tmp, fullSrc, s, c);
+                InsertSlice(tmp, source, 0, 0);
+                InsertSlice(tmp, source, Ls_-1, 0);
+                axpby_ssp_pplus(source, 0., source, 1., source, 0, 0);
+                axpby_ssp_pminus(source, 0., source, 1., source, Ls_-1, Ls_-1);
+            }
+        }
+        // source conversion for 5D sources
+        else
+        {
+            if (Ls_ != env().getObjectLs(par().source))
+            {
+                HADRON_ERROR("Ls mismatch between quark action and source");
+            }
+            else
+            {
+                PropToFerm(source, fullSrc, s, c);
+            }
+        }
+        sol = zero;
+        solver(sol, source);
+        FermToProp(prop, sol, s, c);
+        // create 4D propagators from 5D one if necessary
+        if (Ls_ > 1)
+        {
+            PropagatorField &p4d =
+            *env().template getObject<PropagatorField>(getName());
+            
+            axpby_ssp_pminus(sol, 0., sol, 1., sol, 0, 0);
+            axpby_ssp_pplus(sol, 1., sol, 1., sol, 0, Ls_-1);
+            ExtractSlice(tmp, sol, 0, 0);
+            FermToProp(p4d, tmp, s, c);
+        }
+    }
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MFermion_GaugeProp_hpp_

extras/Hadrons/Modules/MGauge/Load.cc

@@ -0,0 +1,78 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MGauge/Load.cc
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Hadrons/Modules/MGauge/Load.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MGauge;
+
+/******************************************************************************
+*                           TLoad implementation                               *
+******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TLoad::TLoad(const std::string name)
+: Module<LoadPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TLoad::getInput(void)
+{
+    std::vector<std::string> in;
+    
+    return in;
+}
+
+std::vector<std::string> TLoad::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TLoad::setup(void)
+{
+    env().registerLattice<LatticeGaugeField>(getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TLoad::execute(void)
+{
+    FieldMetaData  header;
+    std::string fileName = par().file + "."
+                           + std::to_string(env().getTrajectory());
+    
+    LOG(Message) << "Loading NERSC configuration from file '" << fileName
+                 << "'" << std::endl;
+    LatticeGaugeField &U = *env().createLattice<LatticeGaugeField>(getName());
+    NerscIO::readConfiguration(U, header, fileName);
+    LOG(Message) << "NERSC header:" << std::endl;
+    dump_meta_data(header, LOG(Message));
+}

extras/Hadrons/Modules/MGauge/Load.hpp

@@ -0,0 +1,73 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MGauge/Load.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MGauge_Load_hpp_
+#define Hadrons_MGauge_Load_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         Load a NERSC configuration                         *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MGauge)
+
+class LoadPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(LoadPar,
+                                    std::string, file);
+};
+
+class TLoad: public Module<LoadPar>
+{
+public:
+    // constructor
+    TLoad(const std::string name);
+    // destructor
+    virtual ~TLoad(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(Load, TLoad, MGauge);
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MGauge_Load_hpp_

extras/Hadrons/Modules/MGauge/Random.cc

@@ -0,0 +1,69 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MGauge/Random.cc
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Hadrons/Modules/MGauge/Random.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MGauge;
+
+/******************************************************************************
+*                           TRandom implementation                            *
+******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TRandom::TRandom(const std::string name)
+: Module<NoPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TRandom::getInput(void)
+{
+    return std::vector<std::string>();
+}
+
+std::vector<std::string> TRandom::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TRandom::setup(void)
+{
+    env().registerLattice<LatticeGaugeField>(getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TRandom::execute(void)
+{
+    LOG(Message) << "Generating random gauge configuration" << std::endl;
+    LatticeGaugeField &U = *env().createLattice<LatticeGaugeField>(getName());
+    SU3::HotConfiguration(*env().get4dRng(), U);
+}

extras/Hadrons/Modules/MGauge/Random.hpp

@@ -0,0 +1,66 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MGauge/Random.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MGauge_Random_hpp_
+#define Hadrons_MGauge_Random_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                             Random gauge                                   *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MGauge)
+
+class TRandom: public Module<NoPar>
+{
+public:
+    // constructor
+    TRandom(const std::string name);
+    // destructor
+    virtual ~TRandom(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(Random, TRandom, MGauge);
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MGauge_Random_hpp_

extras/Hadrons/Modules/MGauge/StochEm.cc

@@ -0,0 +1,88 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MGauge/StochEm.cc
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/Hadrons/Modules/MGauge/StochEm.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MGauge;
+
+/******************************************************************************
+*                  TStochEm implementation                             *
+******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TStochEm::TStochEm(const std::string name)
+: Module<StochEmPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TStochEm::getInput(void)
+{
+    std::vector<std::string> in;
+    
+    return in;
+}
+
+std::vector<std::string> TStochEm::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TStochEm::setup(void)
+{
+    if (!env().hasRegisteredObject("_" + getName() + "_weight"))
+    {
+        env().registerLattice<EmComp>("_" + getName() + "_weight");
+    }
+    env().registerLattice<EmField>(getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TStochEm::execute(void)
+{
+    PhotonR photon(par().gauge, par().zmScheme);
+    EmField &a = *env().createLattice<EmField>(getName());
+    EmComp  *w;
+    
+    if (!env().hasCreatedObject("_" + getName() + "_weight"))
+    {
+        LOG(Message) << "Caching stochatic EM potential weight (gauge: "
+                     << par().gauge << ", zero-mode scheme: "
+                     << par().zmScheme << ")..." << std::endl;
+        w = env().createLattice<EmComp>("_" + getName() + "_weight");
+        photon.StochasticWeight(*w);
+    }
+    else
+    {
+        w = env().getObject<EmComp>("_" + getName() + "_weight");
+    }
+    LOG(Message) << "Generating stochatic EM potential..." << std::endl;
+    photon.StochasticField(a, *env().get4dRng(), *w);
+}

extras/Hadrons/Modules/MGauge/StochEm.hpp

@@ -0,0 +1,75 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MGauge/StochEm.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#ifndef Hadrons_MGauge_StochEm_hpp_
+#define Hadrons_MGauge_StochEm_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         StochEm                                 *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MGauge)
+
+class StochEmPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(StochEmPar,
+                                    PhotonR::Gauge,    gauge,
+                                    PhotonR::ZmScheme, zmScheme);
+};
+
+class TStochEm: public Module<StochEmPar>
+{
+public:
+    typedef PhotonR::GaugeField     EmField;
+    typedef PhotonR::GaugeLinkField EmComp;
+public:
+    // constructor
+    TStochEm(const std::string name);
+    // destructor
+    virtual ~TStochEm(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(StochEm, TStochEm, MGauge);
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MGauge_StochEm_hpp_

extras/Hadrons/Modules/MGauge/Unit.cc

@@ -0,0 +1,69 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MGauge/Unit.cc
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Hadrons/Modules/MGauge/Unit.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MGauge;
+
+/******************************************************************************
+*                            TUnit implementation                             *
+******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TUnit::TUnit(const std::string name)
+: Module<NoPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TUnit::getInput(void)
+{
+    return std::vector<std::string>();
+}
+
+std::vector<std::string> TUnit::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TUnit::setup(void)
+{
+    env().registerLattice<LatticeGaugeField>(getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TUnit::execute(void)
+{
+    LOG(Message) << "Creating unit gauge configuration" << std::endl;
+    LatticeGaugeField &U = *env().createLattice<LatticeGaugeField>(getName());
+    SU3::ColdConfiguration(*env().get4dRng(), U);
+}

extras/Hadrons/Modules/MGauge/Unit.hpp

@@ -0,0 +1,66 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MGauge/Unit.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MGauge_Unit_hpp_
+#define Hadrons_MGauge_Unit_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                              Unit gauge                                    *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MGauge)
+
+class TUnit: public Module<NoPar>
+{
+public:
+    // constructor
+    TUnit(const std::string name);
+    // destructor
+    virtual ~TUnit(void) = default;
+    // dependencies/products
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(Unit, TUnit, MGauge);
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MGauge_Unit_hpp_

extras/Hadrons/Modules/MLoop/NoiseLoop.hpp

@@ -0,0 +1,132 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MLoop/NoiseLoop.hpp
+
+Copyright (C) 2016
+
+Author: Andrew Lawson <andrew.lawson1991@gmail.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MLoop_NoiseLoop_hpp_
+#define Hadrons_MLoop_NoiseLoop_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/*
+ 
+ Noise loop propagator
+ -----------------------------
+ * loop_x = q_x * adj(eta_x)
+ 
+ * options:
+ - q = Result of inversion on noise source.
+ - eta = noise source.
+
+ */
+
+
+/******************************************************************************
+ *                         NoiseLoop                                          *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MLoop)
+
+class NoiseLoopPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(NoiseLoopPar,
+                                    std::string, q,
+                                    std::string, eta);
+};
+
+template <typename FImpl>
+class TNoiseLoop: public Module<NoiseLoopPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TNoiseLoop(const std::string name);
+    // destructor
+    virtual ~TNoiseLoop(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(NoiseLoop, TNoiseLoop<FIMPL>, MLoop);
+
+/******************************************************************************
+ *                 TNoiseLoop implementation                                  *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TNoiseLoop<FImpl>::TNoiseLoop(const std::string name)
+: Module<NoiseLoopPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TNoiseLoop<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().q, par().eta};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TNoiseLoop<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TNoiseLoop<FImpl>::setup(void)
+{
+    env().template registerLattice<PropagatorField>(getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TNoiseLoop<FImpl>::execute(void)
+{
+    PropagatorField &loop = *env().template createLattice<PropagatorField>(getName());
+    PropagatorField &q    = *env().template getObject<PropagatorField>(par().q);
+    PropagatorField &eta  = *env().template getObject<PropagatorField>(par().eta);
+    loop = q*adj(eta);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MLoop_NoiseLoop_hpp_

extras/Hadrons/Modules/MScalar/ChargedProp.cc

@@ -0,0 +1,226 @@
+#include <Grid/Hadrons/Modules/MScalar/ChargedProp.hpp>
+#include <Grid/Hadrons/Modules/MScalar/Scalar.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MScalar;
+
+/******************************************************************************
+*                     TChargedProp implementation                             *
+******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TChargedProp::TChargedProp(const std::string name)
+: Module<ChargedPropPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TChargedProp::getInput(void)
+{
+    std::vector<std::string> in = {par().source, par().emField};
+    
+    return in;
+}
+
+std::vector<std::string> TChargedProp::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TChargedProp::setup(void)
+{
+    freeMomPropName_ = FREEMOMPROP(par().mass);
+    phaseName_.clear();
+    for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+    {
+        phaseName_.push_back("_shiftphase_" + std::to_string(mu));
+    }
+    GFSrcName_ = "_" + getName() + "_DinvSrc";
+    if (!env().hasRegisteredObject(freeMomPropName_))
+    {
+        env().registerLattice<ScalarField>(freeMomPropName_);
+    }
+    if (!env().hasRegisteredObject(phaseName_[0]))
+    {
+        for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+        {
+            env().registerLattice<ScalarField>(phaseName_[mu]);
+        }
+    }
+    if (!env().hasRegisteredObject(GFSrcName_))
+    {
+        env().registerLattice<ScalarField>(GFSrcName_);
+    }
+    env().registerLattice<ScalarField>(getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TChargedProp::execute(void)
+{
+    // CACHING ANALYTIC EXPRESSIONS
+    ScalarField &source = *env().getObject<ScalarField>(par().source);
+    Complex     ci(0.0,1.0);
+    FFT         fft(env().getGrid());
+    
+    // cache free scalar propagator
+    if (!env().hasCreatedObject(freeMomPropName_))
+    {
+        LOG(Message) << "Caching momentum space free scalar propagator"
+                     << " (mass= " << par().mass << ")..." << std::endl;
+        freeMomProp_ = env().createLattice<ScalarField>(freeMomPropName_);
+        SIMPL::MomentumSpacePropagator(*freeMomProp_, par().mass);
+    }
+    else
+    {
+        freeMomProp_ = env().getObject<ScalarField>(freeMomPropName_);
+    }
+    // cache G*F*src
+    if (!env().hasCreatedObject(GFSrcName_))
+        
+    {
+        GFSrc_ = env().createLattice<ScalarField>(GFSrcName_);
+        fft.FFT_all_dim(*GFSrc_, source, FFT::forward);
+        *GFSrc_ = (*freeMomProp_)*(*GFSrc_);
+    }
+    else
+    {
+        GFSrc_ = env().getObject<ScalarField>(GFSrcName_);
+    }
+    // cache phases
+    if (!env().hasCreatedObject(phaseName_[0]))
+    {
+        std::vector<int> &l = env().getGrid()->_fdimensions;
+        
+        LOG(Message) << "Caching shift phases..." << std::endl;
+        for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+        {
+            Real    twoPiL = M_PI*2./l[mu];
+            
+            phase_.push_back(env().createLattice<ScalarField>(phaseName_[mu]));
+            LatticeCoordinate(*(phase_[mu]), mu);
+            *(phase_[mu]) = exp(ci*twoPiL*(*(phase_[mu])));
+        }
+    }
+    else
+    {
+        for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+        {
+            phase_.push_back(env().getObject<ScalarField>(phaseName_[mu]));
+        }
+    }
+
+    // PROPAGATOR CALCULATION
+    LOG(Message) << "Computing charged scalar propagator"
+                 << " (mass= " << par().mass
+                 << ", charge= " << par().charge << ")..." << std::endl;
+    
+    ScalarField &prop   = *env().createLattice<ScalarField>(getName());
+    ScalarField buf(env().getGrid());
+    ScalarField &GFSrc = *GFSrc_, &G = *freeMomProp_;
+    double      q = par().charge;
+    
+    // G*F*Src
+    prop = GFSrc;
+
+    // - q*G*momD1*G*F*Src (momD1 = F*D1*Finv)
+    buf = GFSrc;
+    momD1(buf, fft);
+    buf = G*buf;
+    prop = prop - q*buf;
+
+    // + q^2*G*momD1*G*momD1*G*F*Src (here buf = G*momD1*G*F*Src)
+    momD1(buf, fft);
+    prop = prop + q*q*G*buf;
+
+    // - q^2*G*momD2*G*F*Src (momD2 = F*D2*Finv)
+    buf = GFSrc;
+    momD2(buf, fft);
+    prop = prop - q*q*G*buf;
+
+    // final FT
+    fft.FFT_all_dim(prop, prop, FFT::backward);
+    
+    // OUTPUT IF NECESSARY
+    if (!par().output.empty())
+    {
+        std::string           filename = par().output + "." +
+                                         std::to_string(env().getTrajectory());
+        
+        LOG(Message) << "Saving zero-momentum projection to '"
+                     << filename << "'..." << std::endl;
+        
+        CorrWriter            writer(filename);
+        std::vector<TComplex> vecBuf;
+        std::vector<Complex>  result;
+        
+        sliceSum(prop, vecBuf, Tp);
+        result.resize(vecBuf.size());
+        for (unsigned int t = 0; t < vecBuf.size(); ++t)
+        {
+            result[t] = TensorRemove(vecBuf[t]);
+        }
+        write(writer, "charge", q);
+        write(writer, "prop", result);
+    }
+}
+
+void TChargedProp::momD1(ScalarField &s, FFT &fft)
+{
+    EmField     &A = *env().getObject<EmField>(par().emField);
+    ScalarField buf(env().getGrid()), result(env().getGrid()),
+                Amu(env().getGrid());
+    Complex     ci(0.0,1.0);
+
+    result = zero;
+
+    for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+    {
+        Amu = peekLorentz(A, mu);
+        buf = (*phase_[mu])*s;
+        fft.FFT_all_dim(buf, buf, FFT::backward);
+        buf = Amu*buf;
+        fft.FFT_all_dim(buf, buf, FFT::forward);
+        result = result - ci*buf;
+    }
+    fft.FFT_all_dim(s, s, FFT::backward);
+    for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+    {
+        Amu = peekLorentz(A, mu);
+        buf = Amu*s;
+        fft.FFT_all_dim(buf, buf, FFT::forward);
+        result = result + ci*adj(*phase_[mu])*buf;
+    }
+
+    s = result;
+}
+
+void TChargedProp::momD2(ScalarField &s, FFT &fft)
+{
+    EmField     &A = *env().getObject<EmField>(par().emField);
+    ScalarField buf(env().getGrid()), result(env().getGrid()),
+                Amu(env().getGrid());
+
+    result = zero;
+    
+    for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+    {
+        Amu = peekLorentz(A, mu);
+        buf = (*phase_[mu])*s;
+        fft.FFT_all_dim(buf, buf, FFT::backward);
+        buf = Amu*Amu*buf;
+        fft.FFT_all_dim(buf, buf, FFT::forward);
+        result = result + .5*buf;
+    }
+    fft.FFT_all_dim(s, s, FFT::backward);
+    for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+    {
+        Amu = peekLorentz(A, mu);        
+        buf = Amu*Amu*s;
+        fft.FFT_all_dim(buf, buf, FFT::forward);
+        result = result + .5*adj(*phase_[mu])*buf;
+    }
+
+    s = result;
+}

extras/Hadrons/Modules/MScalar/ChargedProp.hpp

@@ -0,0 +1,61 @@
+#ifndef Hadrons_MScalar_ChargedProp_hpp_
+#define Hadrons_MScalar_ChargedProp_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                       Charged scalar propagator                            *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MScalar)
+
+class ChargedPropPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(ChargedPropPar,
+                                    std::string, emField,
+                                    std::string, source,
+                                    double,      mass,
+                                    double,      charge,
+                                    std::string, output);
+};
+
+class TChargedProp: public Module<ChargedPropPar>
+{
+public:
+    SCALAR_TYPE_ALIASES(SIMPL,);
+    typedef PhotonR::GaugeField     EmField;
+    typedef PhotonR::GaugeLinkField EmComp;
+public:
+    // constructor
+    TChargedProp(const std::string name);
+    // destructor
+    virtual ~TChargedProp(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+private:
+    void momD1(ScalarField &s, FFT &fft);
+    void momD2(ScalarField &s, FFT &fft);
+private:
+    std::string                freeMomPropName_, GFSrcName_;
+    std::vector<std::string>   phaseName_;
+    ScalarField                *freeMomProp_, *GFSrc_;
+    std::vector<ScalarField *> phase_;
+    EmField                    *A;
+};
+
+MODULE_REGISTER_NS(ChargedProp, TChargedProp, MScalar);
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MScalar_ChargedProp_hpp_

extras/Hadrons/Modules/MScalar/FreeProp.cc

@@ -0,0 +1,79 @@
+#include <Grid/Hadrons/Modules/MScalar/FreeProp.hpp>
+#include <Grid/Hadrons/Modules/MScalar/Scalar.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MScalar;
+
+/******************************************************************************
+*                        TFreeProp implementation                             *
+******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TFreeProp::TFreeProp(const std::string name)
+: Module<FreePropPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TFreeProp::getInput(void)
+{
+    std::vector<std::string> in = {par().source};
+    
+    return in;
+}
+
+std::vector<std::string> TFreeProp::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TFreeProp::setup(void)
+{
+    freeMomPropName_ = FREEMOMPROP(par().mass);
+    
+    if (!env().hasRegisteredObject(freeMomPropName_))
+    {
+        env().registerLattice<ScalarField>(freeMomPropName_);
+    }
+    env().registerLattice<ScalarField>(getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TFreeProp::execute(void)
+{
+    ScalarField &prop   = *env().createLattice<ScalarField>(getName());
+    ScalarField &source = *env().getObject<ScalarField>(par().source);
+    ScalarField *freeMomProp;
+
+    if (!env().hasCreatedObject(freeMomPropName_))
+    {
+        LOG(Message) << "Caching momentum space free scalar propagator"
+                     << " (mass= " << par().mass << ")..." << std::endl;
+        freeMomProp = env().createLattice<ScalarField>(freeMomPropName_);
+        SIMPL::MomentumSpacePropagator(*freeMomProp, par().mass);
+    }
+    else
+    {
+        freeMomProp = env().getObject<ScalarField>(freeMomPropName_);
+    }
+    LOG(Message) << "Computing free scalar propagator..." << std::endl;
+    SIMPL::FreePropagator(source, prop, *freeMomProp);
+    
+    if (!par().output.empty())
+    {
+        TextWriter            writer(par().output + "." +
+                                     std::to_string(env().getTrajectory()));
+        std::vector<TComplex> buf;
+        std::vector<Complex>  result;
+        
+        sliceSum(prop, buf, Tp);
+        result.resize(buf.size());
+        for (unsigned int t = 0; t < buf.size(); ++t)
+        {
+            result[t] = TensorRemove(buf[t]);
+        }
+        write(writer, "prop", result);
+    }
+}

extras/Hadrons/Modules/MScalar/FreeProp.hpp

@@ -0,0 +1,50 @@
+#ifndef Hadrons_MScalar_FreeProp_hpp_
+#define Hadrons_MScalar_FreeProp_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                               FreeProp                                     *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MScalar)
+
+class FreePropPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(FreePropPar,
+                                    std::string, source,
+                                    double,      mass,
+                                    std::string, output);
+};
+
+class TFreeProp: public Module<FreePropPar>
+{
+public:
+    SCALAR_TYPE_ALIASES(SIMPL,);
+public:
+    // constructor
+    TFreeProp(const std::string name);
+    // destructor
+    virtual ~TFreeProp(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+private:
+    std::string freeMomPropName_;
+};
+
+MODULE_REGISTER_NS(FreeProp, TFreeProp, MScalar);
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MScalar_FreeProp_hpp_

extras/Hadrons/Modules/MScalar/Scalar.hpp

@@ -0,0 +1,6 @@
+#ifndef Hadrons_Scalar_hpp_
+#define Hadrons_Scalar_hpp_
+
+#define FREEMOMPROP(m) "_scalar_mom_prop_" + std::to_string(m)
+
+#endif // Hadrons_Scalar_hpp_

extras/Hadrons/Modules/MSink/Point.hpp

@@ -0,0 +1,114 @@
+#ifndef Hadrons_MSink_Point_hpp_
+#define Hadrons_MSink_Point_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                                   Point                                    *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MSink)
+
+class PointPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(PointPar,
+                                    std::string, mom);
+};
+
+template <typename FImpl>
+class TPoint: public Module<PointPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl,);
+    SINK_TYPE_ALIASES();
+public:
+    // constructor
+    TPoint(const std::string name);
+    // destructor
+    virtual ~TPoint(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(Point,       TPoint<FIMPL>,        MSink);
+MODULE_REGISTER_NS(ScalarPoint, TPoint<ScalarImplCR>, MSink);
+
+/******************************************************************************
+ *                          TPoint implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TPoint<FImpl>::TPoint(const std::string name)
+: Module<PointPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TPoint<FImpl>::getInput(void)
+{
+    std::vector<std::string> in;
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TPoint<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TPoint<FImpl>::setup(void)
+{
+    unsigned int size;
+    
+    size = env().template lattice4dSize<LatticeComplex>();
+    env().registerObject(getName(), size);
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TPoint<FImpl>::execute(void)
+{
+    std::vector<Real> p = strToVec<Real>(par().mom);
+    LatticeComplex    ph(env().getGrid()), coor(env().getGrid());
+    Complex           i(0.0,1.0);
+    
+    LOG(Message) << "Setting up point sink function for momentum ["
+                 << par().mom << "]" << std::endl;
+    ph = zero;
+    for(unsigned int mu = 0; mu < env().getNd(); mu++)
+    {
+        LatticeCoordinate(coor, mu);
+        ph = ph + (p[mu]/env().getGrid()->_fdimensions[mu])*coor;
+    }
+    ph = exp((Real)(2*M_PI)*i*ph);
+    auto sink = [ph](const PropagatorField &field)
+    {
+        SlicedPropagator res;
+        PropagatorField  tmp = ph*field;
+        
+        sliceSum(tmp, res, Tp);
+        
+        return res;
+    };
+    env().setObject(getName(), new SinkFn(sink));
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MSink_Point_hpp_

extras/Hadrons/Modules/MSolver/RBPrecCG.hpp

@@ -0,0 +1,132 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MSolver/RBPrecCG.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MSolver_RBPrecCG_hpp_
+#define Hadrons_MSolver_RBPrecCG_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                     Schur red-black preconditioned CG                      *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MSolver)
+
+class RBPrecCGPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(RBPrecCGPar,
+                                    std::string, action,
+                                    double     , residual);
+};
+
+template <typename FImpl>
+class TRBPrecCG: public Module<RBPrecCGPar>
+{
+public:
+    FGS_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TRBPrecCG(const std::string name);
+    // destructor
+    virtual ~TRBPrecCG(void) = default;
+    // dependencies/products
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(RBPrecCG, TRBPrecCG<FIMPL>, MSolver);
+
+/******************************************************************************
+ *                      TRBPrecCG template implementation                     *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TRBPrecCG<FImpl>::TRBPrecCG(const std::string name)
+: Module(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TRBPrecCG<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().action};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TRBPrecCG<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TRBPrecCG<FImpl>::setup(void)
+{
+    auto Ls = env().getObjectLs(par().action);
+    
+    env().registerObject(getName(), 0, Ls);
+    env().addOwnership(getName(), par().action);
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TRBPrecCG<FImpl>::execute(void)
+{
+    auto &mat   = *(env().template getObject<FMat>(par().action));
+    auto solver = [&mat, this](FermionField &sol, const FermionField &source)
+    {
+        ConjugateGradient<FermionField>           cg(par().residual, 10000);
+        SchurRedBlackDiagMooeeSolve<FermionField> schurSolver(cg);
+        
+        schurSolver(mat, source, sol);
+    };
+    
+    LOG(Message) << "setting up Schur red-black preconditioned CG for"
+                 << " action '" << par().action << "' with residual "
+                 << par().residual << std::endl;
+    env().setObject(getName(), new SolverFn(solver));
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MSolver_RBPrecCG_hpp_

extras/Hadrons/Modules/MSource/Point.hpp

@@ -0,0 +1,136 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MSource/Point.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MSource_Point_hpp_
+#define Hadrons_MSource_Point_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/*
+ 
+ Point source
+ ------------
+ * src_x = delta_x,position
+ 
+ * options:
+ - position: space-separated integer sequence (e.g. "0 1 1 0")
+ 
+ */
+
+/******************************************************************************
+ *                                  TPoint                                     *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MSource)
+
+class PointPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(PointPar,
+                                    std::string, position);
+};
+
+template <typename FImpl>
+class TPoint: public Module<PointPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TPoint(const std::string name);
+    // destructor
+    virtual ~TPoint(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(Point,       TPoint<FIMPL>,        MSource);
+MODULE_REGISTER_NS(ScalarPoint, TPoint<ScalarImplCR>, MSource);
+
+/******************************************************************************
+ *                       TPoint template implementation                       *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TPoint<FImpl>::TPoint(const std::string name)
+: Module<PointPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TPoint<FImpl>::getInput(void)
+{
+    std::vector<std::string> in;
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TPoint<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TPoint<FImpl>::setup(void)
+{
+    env().template registerLattice<PropagatorField>(getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TPoint<FImpl>::execute(void)
+{
+    std::vector<int> position = strToVec<int>(par().position);
+    typename SitePropagator::scalar_object id;
+    
+    LOG(Message) << "Creating point source at position [" << par().position
+                 << "]" << std::endl;
+    PropagatorField &src = *env().template createLattice<PropagatorField>(getName());
+    id  = 1.;
+    src = zero;
+    pokeSite(id, src, position);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MSource_Point_hpp_

extras/Hadrons/Modules/MSource/SeqGamma.hpp

@@ -0,0 +1,164 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MSource/SeqGamma.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+Copyright (C) 2017
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MSource_SeqGamma_hpp_
+#define Hadrons_MSource_SeqGamma_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/*
+ 
+ Sequential source
+ -----------------------------
+ * src_x = q_x * theta(x_3 - tA) * theta(tB - x_3) * gamma * exp(i x.mom)
+ 
+ * options:
+ - q: input propagator (string)
+ - tA: begin timeslice (integer)
+ - tB: end timesilce (integer)
+ - gamma: gamma product to insert (integer)
+ - mom: momentum insertion, space-separated float sequence (e.g ".1 .2 1. 0.")
+ 
+ */
+
+/******************************************************************************
+ *                         SeqGamma                                 *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MSource)
+
+class SeqGammaPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(SeqGammaPar,
+                                    std::string,    q,
+                                    unsigned int,   tA,
+                                    unsigned int,   tB,
+                                    Gamma::Algebra, gamma,
+                                    std::string,    mom);
+};
+
+template <typename FImpl>
+class TSeqGamma: public Module<SeqGammaPar>
+{
+public:
+    FGS_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TSeqGamma(const std::string name);
+    // destructor
+    virtual ~TSeqGamma(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(SeqGamma, TSeqGamma<FIMPL>, MSource);
+
+/******************************************************************************
+ *                         TSeqGamma implementation                           *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TSeqGamma<FImpl>::TSeqGamma(const std::string name)
+: Module<SeqGammaPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TSeqGamma<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().q};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TSeqGamma<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TSeqGamma<FImpl>::setup(void)
+{
+    env().template registerLattice<PropagatorField>(getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TSeqGamma<FImpl>::execute(void)
+{
+    if (par().tA == par().tB)
+    {
+        LOG(Message) << "Generating gamma_" << par().gamma
+                     << " sequential source at t= " << par().tA << std::endl;
+    }
+    else
+    {
+        LOG(Message) << "Generating gamma_" << par().gamma
+                     << " sequential source for "
+                     << par().tA << " <= t <= " << par().tB << std::endl;
+    }
+    PropagatorField &src = *env().template createLattice<PropagatorField>(getName());
+    PropagatorField &q   = *env().template getObject<PropagatorField>(par().q);
+    Lattice<iScalar<vInteger>> t(env().getGrid());
+    LatticeComplex             ph(env().getGrid()), coor(env().getGrid());
+    Gamma                      g(par().gamma);
+    std::vector<Real>          p;
+    Complex                    i(0.0,1.0);
+    
+    p  = strToVec<Real>(par().mom);
+    ph = zero;
+    for(unsigned int mu = 0; mu < env().getNd(); mu++)
+    {
+        LatticeCoordinate(coor, mu);
+        ph = ph + p[mu]*coor*((1./(env().getGrid()->_fdimensions[mu])));
+    }
+    ph = exp((Real)(2*M_PI)*i*ph);
+    LatticeCoordinate(t, Tp);
+    src = where((t >= par().tA) and (t <= par().tB), ph*(g*q), 0.*q);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MSource_SeqGamma_hpp_

extras/Hadrons/Modules/MSource/Wall.hpp

@@ -0,0 +1,147 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MSource/Wall.hpp
+
+Copyright (C) 2017
+
+Author: Andrew Lawson <andrew.lawson1991@gmail.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MSource_WallSource_hpp_
+#define Hadrons_MSource_WallSource_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/*
+ 
+ Wall source
+ -----------------------------
+ * src_x = delta(x_3 - tW) * exp(i x.mom)
+ 
+ * options:
+ - tW: source timeslice (integer)
+ - mom: momentum insertion, space-separated float sequence (e.g ".1 .2 1. 0.")
+ 
+ */
+
+/******************************************************************************
+ *                         Wall                                               *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MSource)
+
+class WallPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(WallPar,
+                                    unsigned int, tW,
+                                    std::string, mom);
+};
+
+template <typename FImpl>
+class TWall: public Module<WallPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TWall(const std::string name);
+    // destructor
+    virtual ~TWall(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(Wall, TWall<FIMPL>, MSource);
+
+/******************************************************************************
+ *                 TWall implementation                                       *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TWall<FImpl>::TWall(const std::string name)
+: Module<WallPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TWall<FImpl>::getInput(void)
+{
+    std::vector<std::string> in;
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TWall<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TWall<FImpl>::setup(void)
+{
+    env().template registerLattice<PropagatorField>(getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TWall<FImpl>::execute(void)
+{    
+    LOG(Message) << "Generating wall source at t = " << par().tW 
+                 << " with momentum " << par().mom << std::endl;
+    
+    PropagatorField &src = *env().template createLattice<PropagatorField>(getName());
+    Lattice<iScalar<vInteger>> t(env().getGrid());
+    LatticeComplex             ph(env().getGrid()), coor(env().getGrid());
+    std::vector<Real>          p;
+    Complex                    i(0.0,1.0);
+    
+    p  = strToVec<Real>(par().mom);
+    ph = zero;
+    for(unsigned int mu = 0; mu < Nd; mu++)
+    {
+        LatticeCoordinate(coor, mu);
+        ph = ph + p[mu]*coor*((1./(env().getGrid()->_fdimensions[mu])));
+    }
+    ph = exp((Real)(2*M_PI)*i*ph);
+    LatticeCoordinate(t, Tp);
+    src = 1.;
+    src = where((t == par().tW), src*ph, 0.*src);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MSource_WallSource_hpp_

extras/Hadrons/Modules/MSource/Z2.hpp

@@ -0,0 +1,152 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MSource/Z2.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef Hadrons_MSource_Z2_hpp_
+#define Hadrons_MSource_Z2_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/*
+ 
+ Z_2 stochastic source
+ -----------------------------
+ * src_x = eta_x * theta(x_3 - tA) * theta(tB - x_3)
+ 
+ the eta_x are independent uniform random numbers in {+/- 1 +/- i}
+ 
+ * options:
+ - tA: begin timeslice (integer)
+ - tB: end timesilce (integer)
+ 
+ */
+ 
+/******************************************************************************
+ *                          Z2 stochastic source                              *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MSource)
+
+class Z2Par: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(Z2Par,
+                                    unsigned int, tA,
+                                    unsigned int, tB);
+};
+
+template <typename FImpl>
+class TZ2: public Module<Z2Par>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TZ2(const std::string name);
+    // destructor
+    virtual ~TZ2(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(Z2,       TZ2<FIMPL>,        MSource);
+MODULE_REGISTER_NS(ScalarZ2, TZ2<ScalarImplCR>, MSource);
+
+/******************************************************************************
+ *                       TZ2 template implementation                          *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TZ2<FImpl>::TZ2(const std::string name)
+: Module<Z2Par>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TZ2<FImpl>::getInput(void)
+{
+    std::vector<std::string> in;
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TZ2<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TZ2<FImpl>::setup(void)
+{
+    env().template registerLattice<PropagatorField>(getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TZ2<FImpl>::execute(void)
+{
+    Lattice<iScalar<vInteger>> t(env().getGrid());
+    LatticeComplex             eta(env().getGrid());
+    Complex                    shift(1., 1.);
+    
+    if (par().tA == par().tB)
+    {
+        LOG(Message) << "Generating Z_2 wall source at t= " << par().tA
+        << std::endl;
+    }
+    else
+    {
+        LOG(Message) << "Generating Z_2 band for " << par().tA << " <= t <= "
+        << par().tB << std::endl;
+    }
+    PropagatorField &src = *env().template createLattice<PropagatorField>(getName());
+    LatticeCoordinate(t, Tp);
+    bernoulli(*env().get4dRng(), eta);
+    eta = (2.*eta - shift)*(1./::sqrt(2.));
+    eta = where((t >= par().tA) and (t <= par().tB), eta, 0.*eta);
+    src = 1.;
+    src = src*eta;
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MSource_Z2_hpp_

extras/Hadrons/Modules/templates/Module.cc.template

@@ -0,0 +1,39 @@
+#include <Grid/Hadrons/Modules/___FILEBASENAME___.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+
+/******************************************************************************
+*                  T___FILEBASENAME___ implementation                             *
+******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+T___FILEBASENAME___::T___FILEBASENAME___(const std::string name)
+: Module<___FILEBASENAME___Par>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> T___FILEBASENAME___::getInput(void)
+{
+    std::vector<std::string> in;
+    
+    return in;
+}
+
+std::vector<std::string> T___FILEBASENAME___::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void T___FILEBASENAME___::setup(void)
+{
+
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void T___FILEBASENAME___::execute(void)
+{
+
+}

extras/Hadrons/Modules/templates/Module.hpp.template

@@ -0,0 +1,40 @@
+#ifndef Hadrons____FILEBASENAME____hpp_
+#define Hadrons____FILEBASENAME____hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         ___FILEBASENAME___                                 *
+ ******************************************************************************/
+class ___FILEBASENAME___Par: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(___FILEBASENAME___Par, 
+                                    unsigned int, i);
+};
+
+class T___FILEBASENAME___: public Module<___FILEBASENAME___Par>
+{
+public:
+    // constructor
+    T___FILEBASENAME___(const std::string name);
+    // destructor
+    virtual ~T___FILEBASENAME___(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER(___FILEBASENAME___, T___FILEBASENAME___);
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons____FILEBASENAME____hpp_

extras/Hadrons/Modules/templates/Module_in_NS.cc.template

@@ -0,0 +1,40 @@
+#include <Grid/Hadrons/Modules/___NAMESPACE___/___FILEBASENAME___.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace ___NAMESPACE___;
+
+/******************************************************************************
+*                  T___FILEBASENAME___ implementation                             *
+******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+T___FILEBASENAME___::T___FILEBASENAME___(const std::string name)
+: Module<___FILEBASENAME___Par>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> T___FILEBASENAME___::getInput(void)
+{
+    std::vector<std::string> in;
+    
+    return in;
+}
+
+std::vector<std::string> T___FILEBASENAME___::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void T___FILEBASENAME___::setup(void)
+{
+
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void T___FILEBASENAME___::execute(void)
+{
+
+}

extras/Hadrons/Modules/templates/Module_in_NS.hpp.template

@@ -0,0 +1,44 @@
+#ifndef Hadrons____NAMESPACE_______FILEBASENAME____hpp_
+#define Hadrons____NAMESPACE_______FILEBASENAME____hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         ___FILEBASENAME___                                 *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(___NAMESPACE___)
+
+class ___FILEBASENAME___Par: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(___FILEBASENAME___Par,
+                                    unsigned int, i);
+};
+
+class T___FILEBASENAME___: public Module<___FILEBASENAME___Par>
+{
+public:
+    // constructor
+    T___FILEBASENAME___(const std::string name);
+    // destructor
+    virtual ~T___FILEBASENAME___(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(___FILEBASENAME___, T___FILEBASENAME___, ___NAMESPACE___);
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons____NAMESPACE_______FILEBASENAME____hpp_

extras/Hadrons/Modules/templates/Module_tmp.hpp.template

@@ -0,0 +1,81 @@
+#ifndef Hadrons____FILEBASENAME____hpp_
+#define Hadrons____FILEBASENAME____hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         ___FILEBASENAME___                                 *
+ ******************************************************************************/
+class ___FILEBASENAME___Par: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(___FILEBASENAME___Par,
+                                    unsigned int, i);
+};
+
+template <typename FImpl>
+class T___FILEBASENAME___: public Module<___FILEBASENAME___Par>
+{
+public:
+    // constructor
+    T___FILEBASENAME___(const std::string name);
+    // destructor
+    virtual ~T___FILEBASENAME___(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER(___FILEBASENAME___, T___FILEBASENAME___<FIMPL>);
+
+/******************************************************************************
+ *                 T___FILEBASENAME___ implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+T___FILEBASENAME___<FImpl>::T___FILEBASENAME___(const std::string name)
+: Module<___FILEBASENAME___Par>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> T___FILEBASENAME___<FImpl>::getInput(void)
+{
+    std::vector<std::string> in;
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> T___FILEBASENAME___<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void T___FILEBASENAME___<FImpl>::setup(void)
+{
+    
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void T___FILEBASENAME___<FImpl>::execute(void)
+{
+    
+}
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons____FILEBASENAME____hpp_

extras/Hadrons/Modules/templates/Module_tmp_in_NS.hpp.template

@@ -0,0 +1,85 @@
+#ifndef Hadrons____NAMESPACE_______FILEBASENAME____hpp_
+#define Hadrons____NAMESPACE_______FILEBASENAME____hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         ___FILEBASENAME___                                 *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(___NAMESPACE___)
+
+class ___FILEBASENAME___Par: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(___FILEBASENAME___Par,
+                                    unsigned int, i);
+};
+
+template <typename FImpl>
+class T___FILEBASENAME___: public Module<___FILEBASENAME___Par>
+{
+public:
+    // constructor
+    T___FILEBASENAME___(const std::string name);
+    // destructor
+    virtual ~T___FILEBASENAME___(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(___FILEBASENAME___, T___FILEBASENAME___<FIMPL>, ___NAMESPACE___);
+
+/******************************************************************************
+ *                 T___FILEBASENAME___ implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+T___FILEBASENAME___<FImpl>::T___FILEBASENAME___(const std::string name)
+: Module<___FILEBASENAME___Par>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> T___FILEBASENAME___<FImpl>::getInput(void)
+{
+    std::vector<std::string> in;
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> T___FILEBASENAME___<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void T___FILEBASENAME___<FImpl>::setup(void)
+{
+    
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void T___FILEBASENAME___<FImpl>::execute(void)
+{
+    
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons____NAMESPACE_______FILEBASENAME____hpp_

extras/Hadrons/add_module.sh

@@ -0,0 +1,31 @@
+#!/usr/bin/env bash
+
+if (( $# != 1 && $# != 2)); then
+    echo "usage: `basename $0` <module name> [<namespace>]" 1>&2
+    exit 1
+fi
+NAME=$1
+NS=$2
+
+if (( $# == 1 )); then
+	if [ -e "Modules/${NAME}.cc" ] || [ -e "Modules/${NAME}.hpp" ]; then
+	    echo "error: files Modules/${NAME}.* already exists" 1>&2
+	    exit 1
+	fi
+	sed "s/___FILEBASENAME___/${NAME}/g" Modules/templates/Module.cc.template > Modules/${NAME}.cc
+	sed "s/___FILEBASENAME___/${NAME}/g" Modules/templates/Module.hpp.template > Modules/${NAME}.hpp
+elif (( $# == 2 )); then
+	mkdir -p Modules/${NS}
+	if [ -e "Modules/${NS}/${NAME}.cc" ] || [ -e "Modules/${NS}/${NAME}.hpp" ]; then
+	    echo "error: files Modules/${NS}/${NAME}.* already exists" 1>&2
+	    exit 1
+	fi
+	TMPCC=".${NS}.${NAME}.tmp.cc"
+	TMPHPP=".${NS}.${NAME}.tmp.hpp"
+	sed "s/___FILEBASENAME___/${NAME}/g" Modules/templates/Module_in_NS.cc.template  > ${TMPCC}
+	sed "s/___FILEBASENAME___/${NAME}/g" Modules/templates/Module_in_NS.hpp.template > ${TMPHPP}
+	sed "s/___NAMESPACE___/${NS}/g" ${TMPCC}  > Modules/${NS}/${NAME}.cc
+	sed "s/___NAMESPACE___/${NS}/g" ${TMPHPP} > Modules/${NS}/${NAME}.hpp
+	rm -f ${TMPCC} ${TMPHPP}
+fi
+./make_module_list.sh

extras/Hadrons/add_module_template.sh

@@ -0,0 +1,28 @@
+#!/usr/bin/env bash
+
+if (( $# != 1 && $# != 2)); then
+    echo "usage: `basename $0` <module name> [<namespace>]" 1>&2
+    exit 1
+fi
+NAME=$1
+NS=$2
+
+if (( $# == 1 )); then
+	if [ -e "Modules/${NAME}.cc" ] || [ -e "Modules/${NAME}.hpp" ]; then
+	    echo "error: files Modules/${NAME}.* already exists" 1>&2
+	    exit 1
+	fi
+	sed "s/___FILEBASENAME___/${NAME}/g" Modules/templates/Module_tmp.hpp.template > Modules/${NAME}.hpp
+elif (( $# == 2 )); then
+	mkdir -p Modules/${NS}
+	if [ -e "Modules/${NS}/${NAME}.cc" ] || [ -e "Modules/${NS}/${NAME}.hpp" ]; then
+	    echo "error: files Modules/${NS}/${NAME}.* already exists" 1>&2
+	    exit 1
+	fi
+	TMPCC=".${NS}.${NAME}.tmp.cc"
+	TMPHPP=".${NS}.${NAME}.tmp.hpp"
+	sed "s/___FILEBASENAME___/${NAME}/g" Modules/templates/Module_tmp_in_NS.hpp.template > ${TMPHPP}
+	sed "s/___NAMESPACE___/${NS}/g" ${TMPHPP} > Modules/${NS}/${NAME}.hpp
+	rm -f ${TMPCC} ${TMPHPP}
+fi
+./make_module_list.sh

extras/Hadrons/make_module_list.sh

@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+
+echo 'modules_cc =\' > modules.inc
+find Modules -name '*.cc' -type f -print | sed 's/^/  /;$q;s/$/ \\/' >> modules.inc
+echo '' >> modules.inc
+echo 'modules_hpp =\' >> modules.inc
+find Modules -name '*.hpp' -type f -print | sed 's/^/  /;$q;s/$/ \\/' >> modules.inc
+echo '' >> modules.inc
+rm -f Modules.hpp
+for f in `find Modules -name '*.hpp'`; do
+	echo "#include <Grid/Hadrons/${f}>" >> Modules.hpp
+done

extras/Hadrons/modules.inc

@@ -0,0 +1,38 @@
+modules_cc =\
+  Modules/MContraction/WeakHamiltonianEye.cc \
+  Modules/MContraction/WeakHamiltonianNonEye.cc \
+  Modules/MContraction/WeakNeutral4ptDisc.cc \
+  Modules/MGauge/Load.cc \
+  Modules/MGauge/Random.cc \
+  Modules/MGauge/StochEm.cc \
+  Modules/MGauge/Unit.cc \
+  Modules/MScalar/ChargedProp.cc \
+  Modules/MScalar/FreeProp.cc
+
+modules_hpp =\
+  Modules/MAction/DWF.hpp \
+  Modules/MAction/Wilson.hpp \
+  Modules/MContraction/Baryon.hpp \
+  Modules/MContraction/DiscLoop.hpp \
+  Modules/MContraction/Gamma3pt.hpp \
+  Modules/MContraction/Meson.hpp \
+  Modules/MContraction/WeakHamiltonian.hpp \
+  Modules/MContraction/WeakHamiltonianEye.hpp \
+  Modules/MContraction/WeakHamiltonianNonEye.hpp \
+  Modules/MContraction/WeakNeutral4ptDisc.hpp \
+  Modules/MFermion/GaugeProp.hpp \
+  Modules/MGauge/Load.hpp \
+  Modules/MGauge/Random.hpp \
+  Modules/MGauge/StochEm.hpp \
+  Modules/MGauge/Unit.hpp \
+  Modules/MLoop/NoiseLoop.hpp \
+  Modules/MScalar/ChargedProp.hpp \
+  Modules/MScalar/FreeProp.hpp \
+  Modules/MScalar/Scalar.hpp \
+  Modules/MSink/Point.hpp \
+  Modules/MSolver/RBPrecCG.hpp \
+  Modules/MSource/Point.hpp \
+  Modules/MSource/SeqGamma.hpp \
+  Modules/MSource/Wall.hpp \
+  Modules/MSource/Z2.hpp
+

extras/Makefile.am

@@ -0,0 +1 @@
+SUBDIRS = Hadrons

extras/qed-fvol/Global.cc

@@ -0,0 +1,11 @@
+#include <qed-fvol/Global.hpp>
+
+using namespace Grid;
+using namespace QCD;
+using namespace QedFVol;
+
+QedFVolLogger QedFVol::QedFVolLogError(1,"Error");
+QedFVolLogger QedFVol::QedFVolLogWarning(1,"Warning");
+QedFVolLogger QedFVol::QedFVolLogMessage(1,"Message");
+QedFVolLogger QedFVol::QedFVolLogIterative(1,"Iterative");
+QedFVolLogger QedFVol::QedFVolLogDebug(1,"Debug");

extras/qed-fvol/Global.hpp

@@ -0,0 +1,42 @@
+#ifndef QedFVol_Global_hpp_
+#define QedFVol_Global_hpp_
+
+#include <Grid/Grid.h>
+
+#define BEGIN_QEDFVOL_NAMESPACE \
+namespace Grid {\
+using namespace QCD;\
+namespace QedFVol {\
+using Grid::operator<<;
+#define END_QEDFVOL_NAMESPACE }}
+
+/* the 'using Grid::operator<<;' statement prevents a very nasty compilation
+ * error with GCC (clang compiles fine without it).
+ */
+
+BEGIN_QEDFVOL_NAMESPACE
+
+class QedFVolLogger: public Logger
+{
+public:
+    QedFVolLogger(int on, std::string nm): Logger("QedFVol", on, nm,
+                                                  GridLogColours, "BLACK"){};
+};
+
+#define LOG(channel) std::cout << QedFVolLog##channel
+#define QEDFVOL_ERROR(msg)\
+LOG(Error) << msg << " (" << __FUNCTION__ << " at " << __FILE__ << ":"\
+           << __LINE__ << ")" << std::endl;\
+abort();
+
+#define DEBUG_VAR(var) LOG(Debug) << #var << "= " << (var) << std::endl;
+
+extern QedFVolLogger QedFVolLogError;
+extern QedFVolLogger QedFVolLogWarning;
+extern QedFVolLogger QedFVolLogMessage;
+extern QedFVolLogger QedFVolLogIterative;
+extern QedFVolLogger QedFVolLogDebug;
+
+END_QEDFVOL_NAMESPACE
+
+#endif // QedFVol_Global_hpp_

extras/qed-fvol/Makefile.am

@@ -0,0 +1,9 @@
+AM_CXXFLAGS += -I$(top_srcdir)/extras
+
+bin_PROGRAMS = qed-fvol
+
+qed_fvol_SOURCES =   \
+    qed-fvol.cc      \
+    Global.cc
+
+qed_fvol_LDADD   = -lGrid

extras/qed-fvol/WilsonLoops.h

@@ -0,0 +1,265 @@
+#ifndef QEDFVOL_WILSONLOOPS_H
+#define QEDFVOL_WILSONLOOPS_H
+
+#include <Global.hpp>
+
+BEGIN_QEDFVOL_NAMESPACE
+
+template <class Gimpl> class NewWilsonLoops : public Gimpl {
+public:
+  INHERIT_GIMPL_TYPES(Gimpl);
+
+  typedef typename Gimpl::GaugeLinkField GaugeMat;
+  typedef typename Gimpl::GaugeField GaugeLorentz;
+
+  //////////////////////////////////////////////////
+  // directed plaquette oriented in mu,nu plane
+  //////////////////////////////////////////////////
+  static void dirPlaquette(GaugeMat &plaq, const std::vector<GaugeMat> &U,
+                           const int mu, const int nu) {
+    // Annoyingly, must use either scope resolution to find dependent base
+    // class,
+    // or this-> ; there is no "this" in a static method. This forces explicit
+    // Gimpl scope
+    // resolution throughout the usage in this file, and rather defeats the
+    // purpose of deriving
+    // from Gimpl.
+    plaq = Gimpl::CovShiftBackward(
+        U[mu], mu, Gimpl::CovShiftBackward(
+                       U[nu], nu, Gimpl::CovShiftForward(U[mu], mu, U[nu])));
+  }
+  //////////////////////////////////////////////////
+  // trace of directed plaquette oriented in mu,nu plane
+  //////////////////////////////////////////////////
+  static void traceDirPlaquette(LatticeComplex &plaq,
+                                const std::vector<GaugeMat> &U, const int mu,
+                                const int nu) {
+    GaugeMat sp(U[0]._grid);
+    dirPlaquette(sp, U, mu, nu);
+    plaq = trace(sp);
+  }
+  //////////////////////////////////////////////////
+  // sum over all planes of plaquette
+  //////////////////////////////////////////////////
+  static void sitePlaquette(LatticeComplex &Plaq,
+                            const std::vector<GaugeMat> &U) {
+    LatticeComplex sitePlaq(U[0]._grid);
+    Plaq = zero;
+    for (int mu = 1; mu < U[0]._grid->_ndimension; mu++) {
+      for (int nu = 0; nu < mu; nu++) {
+        traceDirPlaquette(sitePlaq, U, mu, nu);
+        Plaq = Plaq + sitePlaq;
+      }
+    }
+  }
+  //////////////////////////////////////////////////
+  // sum over all x,y,z,t and over all planes of plaquette
+  //////////////////////////////////////////////////
+  static Real sumPlaquette(const GaugeLorentz &Umu) {
+    std::vector<GaugeMat> U(4, Umu._grid);
+
+    for (int mu = 0; mu < Umu._grid->_ndimension; mu++) {
+      U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
+    }
+
+    LatticeComplex Plaq(Umu._grid);
+
+    sitePlaquette(Plaq, U);
+
+    TComplex Tp = sum(Plaq);
+    Complex p = TensorRemove(Tp);
+    return p.real();
+  }
+  //////////////////////////////////////////////////
+  // average over all x,y,z,t and over all planes of plaquette
+  //////////////////////////////////////////////////
+  static Real avgPlaquette(const GaugeLorentz &Umu) {
+    int ndim = Umu._grid->_ndimension;
+    Real sumplaq = sumPlaquette(Umu);
+    Real vol = Umu._grid->gSites();
+    Real faces = (1.0 * ndim * (ndim - 1)) / 2.0;
+    return sumplaq / vol / faces / Nc; // Nc dependent... FIXME
+  }
+
+  //////////////////////////////////////////////////
+  // Wilson loop of size (R1, R2), oriented in mu,nu plane
+  //////////////////////////////////////////////////
+  static void wilsonLoop(GaugeMat &wl, const std::vector<GaugeMat> &U,
+                           const int Rmu, const int Rnu,
+                           const int mu, const int nu) {
+    wl = U[nu];
+
+    for(int i = 0; i < Rnu-1; i++){
+      wl = Gimpl::CovShiftForward(U[nu], nu, wl);
+    }
+
+    for(int i = 0; i < Rmu; i++){
+      wl = Gimpl::CovShiftForward(U[mu], mu, wl);
+    }
+
+    for(int i = 0; i < Rnu; i++){
+      wl = Gimpl::CovShiftBackward(U[nu], nu, wl);
+    }
+
+    for(int i = 0; i < Rmu; i++){
+      wl = Gimpl::CovShiftBackward(U[mu], mu, wl);
+    }
+  }
+  //////////////////////////////////////////////////
+  // trace of Wilson Loop oriented in mu,nu plane
+  //////////////////////////////////////////////////
+  static void traceWilsonLoop(LatticeComplex &wl,
+                                const std::vector<GaugeMat> &U,
+                                const int Rmu, const int Rnu,
+                                const int mu, const int nu) {
+    GaugeMat sp(U[0]._grid);
+    wilsonLoop(sp, U, Rmu, Rnu, mu, nu);
+    wl = trace(sp);
+  }
+  //////////////////////////////////////////////////
+  // sum over all planes of Wilson loop
+  //////////////////////////////////////////////////
+  static void siteWilsonLoop(LatticeComplex &Wl,
+                            const std::vector<GaugeMat> &U,
+                            const int R1, const int R2) {
+    LatticeComplex siteWl(U[0]._grid);
+    Wl = zero;
+    for (int mu = 1; mu < U[0]._grid->_ndimension; mu++) {
+      for (int nu = 0; nu < mu; nu++) {
+        traceWilsonLoop(siteWl, U, R1, R2, mu, nu);
+        Wl = Wl + siteWl;
+        traceWilsonLoop(siteWl, U, R2, R1, mu, nu);
+        Wl = Wl + siteWl;
+      }
+    }
+  }
+  //////////////////////////////////////////////////
+  // sum over planes of Wilson loop with length R1
+  // in the time direction
+  //////////////////////////////////////////////////
+  static void siteTimelikeWilsonLoop(LatticeComplex &Wl,
+                            const std::vector<GaugeMat> &U,
+                            const int R1, const int R2) {
+    LatticeComplex siteWl(U[0]._grid);
+
+    int ndim = U[0]._grid->_ndimension;
+
+    Wl = zero;
+    for (int nu = 0; nu < ndim - 1; nu++) {
+      traceWilsonLoop(siteWl, U, R1, R2, ndim-1, nu);
+      Wl = Wl + siteWl;
+    }
+  }
+  //////////////////////////////////////////////////
+  // sum Wilson loop over all planes orthogonal to the time direction
+  //////////////////////////////////////////////////
+  static void siteSpatialWilsonLoop(LatticeComplex &Wl,
+                            const std::vector<GaugeMat> &U,
+                            const int R1, const int R2) {
+    LatticeComplex siteWl(U[0]._grid);
+
+    Wl = zero;
+    for (int mu = 1; mu < U[0]._grid->_ndimension - 1; mu++) {
+      for (int nu = 0; nu < mu; nu++) {
+        traceWilsonLoop(siteWl, U, R1, R2, mu, nu);
+        Wl = Wl + siteWl;
+        traceWilsonLoop(siteWl, U, R2, R1, mu, nu);
+        Wl = Wl + siteWl;
+      }
+    }
+  }
+  //////////////////////////////////////////////////
+  // sum over all x,y,z,t and over all planes of Wilson loop
+  //////////////////////////////////////////////////
+  static Real sumWilsonLoop(const GaugeLorentz &Umu,
+                            const int R1, const int R2) {
+    std::vector<GaugeMat> U(4, Umu._grid);
+
+    for (int mu = 0; mu < Umu._grid->_ndimension; mu++) {
+      U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
+    }
+
+    LatticeComplex Wl(Umu._grid);
+
+    siteWilsonLoop(Wl, U, R1, R2);
+
+    TComplex Tp = sum(Wl);
+    Complex p = TensorRemove(Tp);
+    return p.real();
+  }
+  //////////////////////////////////////////////////
+  // sum over all x,y,z,t and over all planes of timelike Wilson loop
+  //////////////////////////////////////////////////
+  static Real sumTimelikeWilsonLoop(const GaugeLorentz &Umu,
+                            const int R1, const int R2) {
+    std::vector<GaugeMat> U(4, Umu._grid);
+
+    for (int mu = 0; mu < Umu._grid->_ndimension; mu++) {
+      U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
+    }
+
+    LatticeComplex Wl(Umu._grid);
+
+    siteTimelikeWilsonLoop(Wl, U, R1, R2);
+
+    TComplex Tp = sum(Wl);
+    Complex p = TensorRemove(Tp);
+    return p.real();
+  }
+  //////////////////////////////////////////////////
+  // sum over all x,y,z,t and over all planes of spatial Wilson loop
+  //////////////////////////////////////////////////
+  static Real sumSpatialWilsonLoop(const GaugeLorentz &Umu,
+                            const int R1, const int R2) {
+    std::vector<GaugeMat> U(4, Umu._grid);
+
+    for (int mu = 0; mu < Umu._grid->_ndimension; mu++) {
+      U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
+    }
+
+    LatticeComplex Wl(Umu._grid);
+
+    siteSpatialWilsonLoop(Wl, U, R1, R2);
+
+    TComplex Tp = sum(Wl);
+    Complex p = TensorRemove(Tp);
+    return p.real();
+  }
+  //////////////////////////////////////////////////
+  // average over all x,y,z,t and over all planes of Wilson loop
+  //////////////////////////////////////////////////
+  static Real avgWilsonLoop(const GaugeLorentz &Umu,
+                            const int R1, const int R2) {
+    int ndim = Umu._grid->_ndimension;
+    Real sumWl = sumWilsonLoop(Umu, R1, R2);
+    Real vol = Umu._grid->gSites();
+    Real faces = 1.0 * ndim * (ndim - 1);
+    return sumWl / vol / faces / Nc; // Nc dependent... FIXME
+  }
+  //////////////////////////////////////////////////
+  // average over all x,y,z,t and over all planes of timelike Wilson loop
+  //////////////////////////////////////////////////
+  static Real avgTimelikeWilsonLoop(const GaugeLorentz &Umu,
+                            const int R1, const int R2) {
+    int ndim = Umu._grid->_ndimension;
+    Real sumWl = sumTimelikeWilsonLoop(Umu, R1, R2);
+    Real vol = Umu._grid->gSites();
+    Real faces = 1.0 * (ndim - 1);
+    return sumWl / vol / faces / Nc; // Nc dependent... FIXME
+  }
+  //////////////////////////////////////////////////
+  // average over all x,y,z,t and over all planes of spatial Wilson loop
+  //////////////////////////////////////////////////
+  static Real avgSpatialWilsonLoop(const GaugeLorentz &Umu,
+                            const int R1, const int R2) {
+    int ndim = Umu._grid->_ndimension;
+    Real sumWl = sumSpatialWilsonLoop(Umu, R1, R2);
+    Real vol = Umu._grid->gSites();
+    Real faces = 1.0 * (ndim - 1) * (ndim - 2);
+    return sumWl / vol / faces / Nc; // Nc dependent... FIXME
+  }
+};
+
+END_QEDFVOL_NAMESPACE
+
+#endif // QEDFVOL_WILSONLOOPS_H

extras/qed-fvol/qed-fvol.cc

@@ -0,0 +1,88 @@
+#include <Global.hpp>
+#include <WilsonLoops.h>
+
+using namespace Grid;
+using namespace QCD;
+using namespace QedFVol;
+
+typedef PeriodicGaugeImpl<QedGimplR>    QedPeriodicGimplR;
+typedef PhotonR::GaugeField             EmField;
+typedef PhotonR::GaugeLinkField         EmComp;
+
+const int NCONFIGS = 10;
+const int NWILSON = 10;
+
+int main(int argc, char *argv[])
+{
+    // parse command line
+    std::string parameterFileName;
+    
+    if (argc < 2)
+    {
+        std::cerr << "usage: " << argv[0] << " <parameter file> [Grid options]";
+        std::cerr << std::endl;
+        std::exit(EXIT_FAILURE);
+    }
+    parameterFileName = argv[1];
+    
+    // initialization
+    Grid_init(&argc, &argv);
+    QedFVolLogError.Active(GridLogError.isActive());
+    QedFVolLogWarning.Active(GridLogWarning.isActive());
+    QedFVolLogMessage.Active(GridLogMessage.isActive());
+    QedFVolLogIterative.Active(GridLogIterative.isActive());
+    QedFVolLogDebug.Active(GridLogDebug.isActive());
+    LOG(Message) << "Grid initialized" << std::endl;
+    
+    // QED stuff
+    std::vector<int> latt_size   = GridDefaultLatt();
+    std::vector<int> simd_layout = GridDefaultSimd(4, vComplex::Nsimd());
+    std::vector<int> mpi_layout  = GridDefaultMpi();
+    GridCartesian    grid(latt_size,simd_layout,mpi_layout);
+    GridParallelRNG  pRNG(&grid);
+    PhotonR          photon(PhotonR::Gauge::feynman,
+                            PhotonR::ZmScheme::qedL);
+    EmField          a(&grid);
+    EmField          expA(&grid);
+
+    Complex imag_unit(0, 1);
+
+    Real wlA;
+    std::vector<Real> logWlAvg(NWILSON, 0.0), logWlTime(NWILSON, 0.0), logWlSpace(NWILSON, 0.0);
+
+    pRNG.SeedRandomDevice();
+
+    LOG(Message) << "Wilson loop calculation beginning" << std::endl;
+    for(int ic = 0; ic < NCONFIGS; ic++){
+        LOG(Message) << "Configuration " << ic <<std::endl;
+        photon.StochasticField(a, pRNG);
+
+        // Exponentiate photon field
+        expA = exp(imag_unit*a);
+
+        // Calculate Wilson loops
+        for(int iw=1; iw<=NWILSON; iw++){
+            wlA = NewWilsonLoops<QedPeriodicGimplR>::avgWilsonLoop(expA, iw, iw) * 3;
+            logWlAvg[iw-1] -= 2*log(wlA);
+            wlA = NewWilsonLoops<QedPeriodicGimplR>::avgTimelikeWilsonLoop(expA, iw, iw) * 3;
+            logWlTime[iw-1] -= 2*log(wlA);
+            wlA = NewWilsonLoops<QedPeriodicGimplR>::avgSpatialWilsonLoop(expA, iw, iw) * 3;
+            logWlSpace[iw-1] -= 2*log(wlA);
+        }
+    }
+    LOG(Message) << "Wilson loop calculation completed" << std::endl;
+    
+    // Calculate Wilson loops
+    for(int iw=1; iw<=10; iw++){
+        LOG(Message) << iw << 'x' << iw << " Wilson loop" << std::endl;
+        LOG(Message) << "-2log(W) average: " << logWlAvg[iw-1]/NCONFIGS << std::endl;
+        LOG(Message) << "-2log(W) timelike: " << logWlTime[iw-1]/NCONFIGS << std::endl;
+        LOG(Message) << "-2log(W) spatial: " << logWlSpace[iw-1]/NCONFIGS << std::endl;
+    }
+
+    // epilogue
+    LOG(Message) << "Grid is finalizing now" << std::endl;
+    Grid_finalize();
+    
+    return EXIT_SUCCESS;
+}

gcc-bug-report/README

@@ -20,4 +20,17 @@ The simple testcase in this directory is the submitted bug report that encapsula
 problem. The test case works with icpc and with clang++, but fails consistently on g++
 current variants.
 
-Peter
+Peter
+
+
+************
+
+Second GCC bug reported, see Issue 100.
+
+https://wandbox.org/permlink/tzssJza6R9XnqANw
+https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80652
+
+Getting Travis fails under gcc-5 for Test_simd, now that I added more comprehensive testing to the
+CI test suite. The limitations of Travis runtime limits & weak cores are being shown.
+
+Travis uses 5.4.1 for g++-5.

grid-config.in

@@ -0,0 +1,86 @@
+#! /bin/sh
+
+prefix=@prefix@
+exec_prefix=@exec_prefix@
+includedir=@includedir@
+
+usage()
+{
+  cat <<EOF
+Usage: grid-config [OPTION]
+
+Known values for OPTION are:
+
+  --prefix     show Grid installation prefix
+  --cxxflags   print pre-processor and compiler flags
+  --ldflags    print library linking flags
+  --libs       print library linking information
+  --summary    print full build summary
+  --help       display this help and exit
+  --version    output version information
+  --git        print git revision
+
+EOF
+  
+  exit $1
+}
+
+if test $# -eq 0; then
+  usage 1
+fi
+
+cflags=false
+libs=false
+
+while test $# -gt 0; do
+  case "$1" in
+    -*=*) optarg=`echo "$1" | sed 's/[-_a-zA-Z0-9]*=//'` ;;
+    *) optarg= ;;
+  esac
+  
+  case "$1" in
+    --prefix)
+      echo $prefix
+    ;;
+    
+    --version)
+      echo @VERSION@
+      exit 0
+    ;;
+    
+    --git)
+      echo "@GRID_BRANCH@ @GRID_SHA@"
+      exit 0
+    ;;
+    
+    --help)
+      usage 0
+    ;;
+    
+    --cxxflags)
+      echo @GRID_CXXFLAGS@
+    ;;
+    
+    --ldflags)
+      echo @GRID_LDFLAGS@
+    ;;
+    
+    --libs)
+      echo @GRID_LIBS@
+    ;;
+    
+    --summary)
+      echo ""
+      echo "@GRID_SUMMARY@"
+      echo ""
+    ;;
+    
+    *)
+      usage
+      exit 1
+    ;;
+  esac
+  shift
+done
+
+exit 0

lib/AlignedAllocator.h

@@ -1,149 +0,0 @@
-    /*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-    Source file: ./lib/AlignedAllocator.h
-
-    Copyright (C) 2015
-
-Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef GRID_ALIGNED_ALLOCATOR_H
-#define GRID_ALIGNED_ALLOCATOR_H
-
-#ifdef HAVE_MALLOC_MALLOC_H
-#include <malloc/malloc.h>
-#endif
-#ifdef HAVE_MALLOC_H
-#include <malloc.h>
-#endif
-
-#ifdef HAVE_MM_MALLOC_H
-#include <mm_malloc.h>
-#endif
-
-#ifdef GRID_COMMS_SHMEM
-extern "C" { 
-#include <mpp/shmem.h>
-extern void * shmem_align(size_t, size_t);
-extern void  shmem_free(void *);
-}
-#endif
-
-namespace Grid {
-
-////////////////////////////////////////////////////////////////////
-// A lattice of something, but assume the something is SIMDized.
-////////////////////////////////////////////////////////////////////
-template<typename _Tp>
-class alignedAllocator {
-public: 
-  typedef std::size_t     size_type;
-  typedef std::ptrdiff_t  difference_type;
-  typedef _Tp*       pointer;
-  typedef const _Tp* const_pointer;
-  typedef _Tp&       reference;
-  typedef const _Tp& const_reference;
-  typedef _Tp        value_type;
-
-  template<typename _Tp1>  struct rebind { typedef alignedAllocator<_Tp1> other; };
-
-  alignedAllocator() throw() { }
-
-  alignedAllocator(const alignedAllocator&) throw() { }
-
-  template<typename _Tp1> alignedAllocator(const alignedAllocator<_Tp1>&) throw() { }
-
-  ~alignedAllocator() throw() { }
-
-  pointer       address(reference __x)       const { return &__x; }
-  //  const_pointer address(const_reference __x) const { return &__x; }
-
-  size_type  max_size() const throw() { return size_t(-1) / sizeof(_Tp); }
-
-  pointer allocate(size_type __n, const void* _p= 0)
-  { 
-#ifdef GRID_COMMS_SHMEM
-
-    _Tp *ptr = (_Tp *) shmem_align(__n*sizeof(_Tp),64);
-
-
-#define PARANOID_SYMMETRIC_HEAP
-#ifdef PARANOID_SYMMETRIC_HEAP
-    static void * bcast;
-    static long  psync[_SHMEM_REDUCE_SYNC_SIZE];
-
-    bcast = (void *) ptr;
-    shmem_broadcast32((void *)&bcast,(void *)&bcast,sizeof(void *)/4,0,0,0,shmem_n_pes(),psync);
-
-    if ( bcast != ptr ) {
-      std::printf("inconsistent alloc pe %d %lx %lx \n",shmem_my_pe(),bcast,ptr);std::fflush(stdout);
-      BACKTRACEFILE();
-      exit(0);
-    }
-
-    assert( bcast == (void *) ptr);
-
-#endif 
-#else
-
-#ifdef HAVE_MM_MALLOC_H
-    _Tp * ptr = (_Tp *) _mm_malloc(__n*sizeof(_Tp),128);
-#else
-    _Tp * ptr = (_Tp *) memalign(128,__n*sizeof(_Tp));
-#endif
-
-#endif
-    _Tp tmp;
-#ifdef GRID_NUMA
-#pragma omp parallel for schedule(static)
-  for(int i=0;i<__n;i++){
-    ptr[i]=tmp;
-  }
-#endif 
-    return ptr;
-  }
-
-  void deallocate(pointer __p, size_type) { 
-#ifdef GRID_COMMS_SHMEM
-    shmem_free((void *)__p);
-#else
-#ifdef HAVE_MM_MALLOC_H
-    _mm_free((void *)__p); 
-#else
-    free((void *)__p);
-#endif
-#endif
-  }
-  void construct(pointer __p, const _Tp& __val) { };
-  void construct(pointer __p) { };
-
-  void destroy(pointer __p) { };
-};
-
-template<typename _Tp>  inline bool
-operator==(const alignedAllocator<_Tp>&, const alignedAllocator<_Tp>&){ return true; }
-
-template<typename _Tp>  inline bool
-operator!=(const alignedAllocator<_Tp>&, const alignedAllocator<_Tp>&){ return false; }
-    
-}; // namespace Grid
-#endif

lib/DisableWarnings.h

@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/DisableWarnings.h
+
+Copyright (C) 2016
+
+Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#ifndef DISABLE_WARNINGS_H
+#define DISABLE_WARNINGS_H
+
+ //disables and intel compiler specific warning (in json.hpp)
+#pragma warning disable 488  
+
+
+#endif